JP7442134B2 - Component mounting method and component mounting system - Google Patents

Description

The present invention relates to a component mounting method and a component mounting system for mounting components on a board.

Conventionally, an anisotropic conductive film (ACF), which is an anisotropic conductive material, is attached to the edge of a substrate such as a display panel, and electronic components such as a drive circuit are mounted on the part of the substrate where the ACF is attached. 2. Description of the Related Art There is a component mounting system that thermocompresses electronic components onto a board on which electronic components are mounted (for example, see Patent Document 1).

As an example of this type of component mounting system, Patent Document 2 discloses that before the above-mentioned ACF pasting, electronic component mounting, and thermocompression bonding processes are performed, alignment marks provided on the board are imaged. It is disclosed that the position adjustment is performed, and that the ACF attached to the substrate is imaged to determine whether the ACF is attached or not.

Japanese Patent Application Publication No. 2016-186966 Japanese Patent Application Publication No. 2009-71069

For example, in a component mounting system, the quality of the imaged object is determined based on images of the imaged object such as alignment marks and ACFs, but sometimes the imaged object is incorrectly determined to be defective even though it is not defective. . In that case, there is a problem that the operation of the component mounting system stops and productivity decreases.

The present invention provides a component mounting method and the like that can suppress erroneous judgments when determining the quality of an imaging target and suppress a decrease in productivity.

A component mounting method according to an aspect of the present invention includes an image acquisition step of acquiring an image of an imaging target provided on a board, and an image capturing process of the imaging target based on a detected value obtained from the image and a preset reference value. a first determination step of performing a pass/fail determination; a correction step of correcting the determination result of the first determination step to good when it is incorrectly determined as defective in the first determination step; and the imaging target determined to be good. Learning data includes a first mode production process including a mounting process for attaching parts to a board, the image acquired in the first mode production process, the judgment result of the first judgment process, and the correction result of the correction process. a learning model generation step of generating a learning model for determining the quality of the imaging target by machine learning, and a learning model evaluation step of evaluating the proficiency of the learning model generated in the learning model generation step. , when it is evaluated that the proficiency level of the learning model has reached a prescribed level in the learning model evaluation step, the production process of the first mode is performed to determine the quality of the imaging target using the learning model. and a change step of changing to a second mode production process having a second determination step.

A component mounting system according to one aspect of the present invention includes: an image acquisition unit that acquires an image of an imaging target provided on a board; a determining unit that determines the quality of the image; a correcting unit that corrects the determination result of the determining unit to good when the determining unit erroneously determines that the image is defective; A component mounting system comprising: a mounting section for attaching a component to a board having an object; The determination unit further includes a learning model generation unit that generates a learning model for determining the quality of a target, and a learning model evaluation unit that evaluates a proficiency level of the learning model generated by the learning model generation unit. When the learning model evaluation unit evaluates that the proficiency level of the learning model has reached a prescribed level, the learning model is used to determine the quality of the imaging target.

Note that these comprehensive or specific aspects may be realized by a system, a method, an integrated circuit, a computer program, or a computer-readable recording medium such as a CD-ROM. and a recording medium may be used in any combination.

According to the component mounting method and the like of the present invention, it is possible to suppress erroneous judgments when determining the quality of an imaging target, and to suppress a decrease in productivity.

FIG. 1 is a plan view showing a component mounting system according to an embodiment. FIG. 2 is a diagram showing a board handled by each device of the component mounting system according to the embodiment. FIG. 3 is a diagram showing a pasting section of the component mounting system according to the embodiment. FIG. 4 is a diagram showing the main crimping section of the component mounting system according to the embodiment. FIG. 5 is a block diagram showing the functional configuration of the computer of the component mounting system according to the embodiment. FIG. 6 is a diagram showing an image of an alignment mark provided on a substrate, a determination result of the quality of the alignment mark, etc. FIG. 7 is a diagram showing an image of an ACF provided on a board, a determination result of the quality of the ACF, etc. FIG. 8 is a diagram illustrating an example of steps included in the first mode production process of the component mounting method according to the embodiment. FIG. 9 is a flowchart illustrating an example of generating a first learning model based on data of a first mode production process and changing the first mode production process to a second mode production process. FIG. 10 is a flowchart illustrating an example of the production process in the second mode of the component mounting method according to the embodiment. FIG. 11 is a flowchart showing another example of the steps included in the first mode production process. FIG. 12 is a flowchart showing an example of generating a second learning model based on the data of the first mode production process shown in FIG. 11 and changing the first mode production process to the second mode production process. . FIG. 13 is a flowchart showing an example of the second mode production process shown in FIG. 12. FIG. 14 is a flowchart showing the production process of the second mode of the component mounting method according to the first modification of the embodiment. FIG. 15 is a flowchart showing a second mode production process of the component mounting method according to the second modification of the embodiment. FIG. 16 is a flowchart illustrating an example of generating a learning model in the production process of the second mode of the component mounting method according to the third modification of the embodiment.

Embodiments of the present invention will be described in detail below with reference to the drawings. Note that all of the embodiments described below are specific examples of the present invention. Therefore, the numerical values, shapes, materials, components, arrangement and connection forms of the components, steps and order of steps, etc. shown in the following embodiments are merely examples, and are not intended to limit the present invention. Therefore, among the constituent elements in the following embodiments, constituent elements that are not described in the independent claims representing the most significant concept of the present invention will be explained as arbitrary constituent elements.

Furthermore, each figure is a schematic diagram and is not necessarily strictly illustrated. Moreover, in each figure, the same reference numerals are attached to the same constituent members.

Furthermore, in this specification and the drawings, the X-axis, Y-axis, and Z-axis represent three axes of a three-dimensional orthogonal coordinate system. The X-axis and the Y-axis are orthogonal to each other, and both are orthogonal to the Z-axis. Furthermore, in the following embodiments, the substrate transport direction is sometimes described as the X-axis positive direction, the Z-axis positive direction as upward, and the Z-axis negative direction as downward.

(Embodiment)
[1. Schematic configuration of component mounting system]
First, a schematic configuration of a component mounting system according to an embodiment will be described with reference to FIGS. 1 and 2.

FIG. 1 is a plan view showing a component mounting system 1 according to an embodiment. FIG. 2 is a diagram showing the board 3 handled by each device of the component mounting system 1.

As shown in FIG. 1, the component mounting system 1 includes a board loading section 10, a pasting section 20, a temporary crimping section 30, a main crimping section 40, a board unloading section 50, a transport section 60, and a computer 70. , is provided. The substrate carrying-in section 10, the pasting section 20, the temporary pressure bonding section 30, the main pressure bonding section 40, and the substrate carrying-out section 50 are connected in this order.

In FIG. 1, computer 70 is illustrated as functional blocks. The computer 70 is connected to each device, such as the pasting section 20, the temporary crimping section 30, the main crimping section 40, etc., in a wireless communicable manner or in a wired communicable manner via a control line or the like, and controls each device. In this embodiment, the computer 70 is used to generate a learning model for determining the quality of an alignment mark or an ACF (Anisotropic Conductive Film) to be imaged. Generation of the learning model will be described later.

The component mounting system 1 is a system for mounting components on a substrate 3, and is used to produce liquid crystal panels and the like. Note that an example of the component is the ACF 6 or the electronic component 5, which will be described later.

As shown in FIG. 2, in the component mounting system 1, the board 3 is first carried into the board carrying section 10. The substrate 3 is a rectangular substrate, and a plurality of alignment marks M (hereinafter sometimes referred to as marks M) and a plurality of electrode parts 4 are provided on the periphery of the substrate 3. Each of the plurality of electrode parts 4 is formed of a predetermined pattern electrode. The substrate 3 carried into the substrate carrying section 10 is conveyed to the pasting section 20.

In the pasting section 20, first, a plurality of marks M are imaged, and the position of the substrate 3 is adjusted based on the imaged image of each mark M. Next, the ACF 6, which is an anisotropic conductive member, is attached to the region of the substrate 3 where the electrode portion 4 is formed. Further, the ACF 6 attached to the substrate 3 is imaged, and the quality of the attached state of the ACF 6 is determined based on the imaged image of the ACF 6. The substrate 3 on which the ACF 6 has been determined to be in a good state is transported to the temporary pressure bonding section 30 .

In the temporary pressure bonding section 30, first, a plurality of marks M are imaged, and the position of the substrate 3 is adjusted based on the imaged image of each mark M. Next, electronic components 5 such as a drive circuit are temporarily pressed onto the substrate 3 via the ACF 6. The board 3 on which the electronic component 5 is temporarily crimped is transported to the final crimping section 40 .

In the main crimping section 40, first, a plurality of marks M are imaged, and the position of the substrate 3 is adjusted based on the imaged image of each mark M. Next, the electronic component 5 is fully crimped onto the substrate 3 via the ACF 6, and the electronic component 5 is electrically connected to the electrode section 4 of the substrate 3. The board 3 to which the electronic component 5 is connected is transported to the board unloading section 50 and further unloaded from the board unloading section 50.

[2. Configuration of each device of component mounting system]
Next, the configuration of each device of the component mounting system 1 will be explained with reference to FIGS. 1, 3, and 4. As described above, the component mounting system 1 includes the board loading section 10, the pasting section 20, the temporary crimping section 30, the main crimping section 40, the board unloading section 50, the transport section 60, and the computer 70. Be prepared.

As shown in FIG. 1, the substrate loading section 10 includes a stage 11 provided on a base 1a. The substrate 3 is loaded onto the stage 11, and is transported by the transport section 60 to the pasting section 20, the temporary pressure bonding section 30, the main pressure bonding section 40, and the substrate unloading section 50 in this order.

The transfer unit 60 includes a substrate transfer mechanism 62A, a substrate transfer mechanism 62B, and a substrate transfer mechanism 62C on a movable base 61 extending in the X-axis direction across a base 1a, a base 1b, and a base 1c, in order from the upstream side. , and a substrate transport mechanism 62D. Each of the substrate transport mechanisms 62A to 62D includes a base 63 and two arm units 64.

For example, the substrate transport mechanism 62A receives the substrate 3 carried into the stage 11 of the substrate carry-in section 10, and delivers it to the stage 23 of the pasting section 20. The substrate transport mechanism 62B receives the substrate 3 from the stage 23 of the pasting section 20 and delivers it to the stage 33 of the temporary pressure bonding section 30. The substrate transport mechanism 62C receives the substrate 3 from the stage 33 of the temporary pressure bonding section 30 and delivers it to the stage 43 of the main pressure bonding section 40. The substrate transport mechanism 62D receives the substrate 3 from the stage 43 of the main compression bonding section 40 and delivers it to the stage 51 of the substrate unloading section 50.

FIG. 3 is a diagram showing the pasting section 20. As shown in FIG. 3A, the pasting section 20 includes a stage moving section 21, a pasting mechanism 22, and an imaging section 29.

The stage moving unit 21 is a mechanism that moves the substrate 3. The stage moving unit 21 includes a multi-axis table movable in, for example, the X-axis direction, the Y-axis direction, the Z-axis direction, and the rotation direction around the Z-axis, and moves the substrate 3 placed on the stage 23.

The imaging unit 29 is, for example, a camera having an imaging element, and images the mark M provided on the substrate 3. Further, the imaging unit 29 images the ACF 6 after being attached to the substrate 3. When the substrate 3 is transparent, the imaging unit 29 may be arranged below the substrate 3 to image the mark M or the ACF 6 from below the substrate 3. The image captured by the imaging unit 29 is transmitted to the computer 70. Note that the alignment mark M and the ACF 6 are examples of objects to be imaged in this embodiment.

The pasting mechanism 22 includes two pasting heads 25 arranged in the X-axis direction on the front surface of a beam 24 extending in the X-axis direction above the base 1b. Each pasting head 25 has a tape supply section 25a and a pasting tool 25b. Furthermore, a pasting support base 26 that supports the substrate 3 from below is provided at a position below the pasting head 25 .

As shown in FIG. 3B, the stage moving section 21 moves the substrate 3 so that the electrode section 4 of the substrate 3 is located between the pasting head 25 and the pasting support base 26. As shown in FIG. Each pasting head 25 cuts the ACF tape 6a supplied by the tape supply section 25a to match the length of the electrode section 4. The pasting mechanism 22 places the ACF 6 formed by cutting above the electrode section 4 and presses the ACF 6 against the substrate 3 by lowering the pasting tool 25b. As a result, the ACF 6 is attached to the substrate 3.

As shown in FIG. 1, the temporary pressure bonding section 30 includes a stage moving section 31, a component mounting mechanism 32, a component supply section 34, and an imaging section 39.

The stage moving unit 31 includes a multi-axis table similar to the stage moving unit 21, and moves the substrate 3 placed on the stage 33.

The imaging unit 39 is, for example, a camera having an imaging element, and images the mark M provided on the substrate 3. The image captured by the imaging unit 39 is transmitted to the computer 70.

The component supply unit 34 supplies the electronic component 5 to the component mounting mechanism 32.

The component mounting mechanism 32 is provided on the base 1b and includes a mounting head, a mounting head moving mechanism that moves the mounting head, and a mounting support base that supports the substrate 3 from below (not shown). The component mounting mechanism 32 picks up the electronic component 5 supplied by the component supply section 34, and places and presses the electronic component 5 onto the ACF 6. Thereby, the electronic component 5 is temporarily pressure-bonded to the substrate 3.

FIG. 4 is a diagram showing the main crimping section 40. As shown in FIG. The main crimping section 40 includes a stage moving section 41, a crimping mechanism 42, and an imaging section 49.

The stage moving unit 41 includes a multi-axis table movable in the X-axis direction, Y-axis direction, Z-axis direction, and rotation direction around the Z-axis, and moves the substrate 3 placed on the stage 43.

The imaging unit 39 is, for example, a camera having an imaging element, and images the mark M provided on the substrate 3. The image captured by the imaging unit 39 is transmitted to the computer 70.

The crimping mechanism 42 includes a plurality of crimping units 45 and a crimping support section 44 that supports the substrate 3 from below. The plurality of crimping units 45 are arranged in a line above the crimping support section 44 . The crimping unit 45 includes a pressure mechanism 47 and a crimping head 48 that includes a heater. The pressure mechanism 47 has a rod that can be vertically projected and retracted, and a crimping head 48 is provided at the lower end of the rod.

The pressure bonding head 48 is lowered by the drive of the pressure mechanism 47 and presses the electronic component 5 mounted on the substrate 3 while heating it. The ACF 6 is cured by this heating, and the electronic component 5 is finally pressure-bonded to the substrate 3.

The substrate unloading section 50 shown in FIG. 1 includes a stage 51 provided on a base 1c. The substrate 3 transferred from the main compression bonding section 40 is held on a stage 51. The substrate 3 held by the substrate unloading unit 50 is either unloaded to another device for next processing, or taken out from the stage 51 by an operator.

[3. Functional configuration of computer]
Next, the functional configuration of the computer 70 of the component mounting system 1 will be explained with reference to FIG.

FIG. 5 is a block diagram showing the functional configuration of the computer 70 of the component mounting system 1.

The computer 70 is a control device for controlling the operation of each device included in the component mounting system 1. The computer 70 includes a control section 70a and a storage section 70b.

The control section 70a controls the pasting section 20, the temporary pressure bonding section 30, the main pressure bonding section 40, and the conveying section 60. The control unit 70a is realized by, for example, a control program for controlling each device of the component mounting system 1, and a CPU (Central Processing Unit) that executes the control program.

The storage unit 70b stores various data necessary for component attachment, such as the size of the board 3, the type of component to be attached to the board 3, the attachment position, the attachment direction, and the timing for transporting the substrate 3 between each processing section, as well as control. The control program and the like executed by the unit 70a are stored. The storage unit 70b also stores a learning model, which will be described later, learning data necessary to generate the learning model, an evaluation level necessary to evaluate the learning model, and the like. This storage unit 70b is realized by a HDD (Hard Disk Drive), an SSD (Solid State Drive), a ROM (Read Only Memory), a RAM (Random Access Memory), or the like.

In the component mounting system 1, in order to suppress erroneous determination when determining the quality of the imaging target (alignment mark M and ACF 6), a learning model is used to perform the quality determination. In this embodiment, the computer 70 is used to generate and evaluate a learning model.

In order to generate and evaluate a learning model, the computer 70 includes an image acquisition section 71, a determination section 73, a display section 79, a modification input section 72, a modification section 74, a learning model generation section 75, a learning model evaluation section 76, and a modification section. It has a functional configuration such as 77.

The image acquisition section 71 acquires images transmitted from each of the imaging section 29 of the pasting section 20, the imaging section 39 of the temporary compression bonding section 30, and the imaging section 49 of the main compression bonding section 40. The display unit 79 displays the images transmitted from the imaging units 29, 39, and 49 on the screen.

The determination unit 73 determines the quality of the imaging target based on the image acquired by the image acquisition unit 71. Specifically, the determination unit 73 determines the quality of the imaging target based on the detected value obtained from the image and a preset reference value. Examples of the detected values include the position, area, shape, color, etc. of the imaging target within the imaging range (imaging area). The preset reference value is stored in the storage unit 70b as a value corresponding to each detected value.

Here, an example will be described in which the quality of the mark M is determined using the mark M as the imaging target.

FIG. 6 is a diagram showing an image of the alignment mark M provided on the substrate 3 and a determination result of the quality of the alignment mark M. FIG. 6 shows an example in which the mark M has a plus (+) shape so that the center coordinates of the mark M can be easily identified. Note that the mark M is not limited to the plus-shaped shape, but may be square or diamond-shaped.

FIG. 6A shows an example in which the shape and position of the image of the mark M are within the range of the reference values, and the determination result of the mark M is good. FIGS. 6(b) and 6(c) show an example in which the shape of the image of the mark M is outside the reference value range and the pass/fail determination result is poor. FIG. 6(d) shows an example in which the position of the image of the mark M is outside the reference value range and the quality determination result is poor.

As shown in FIG. 6(a), if the determination result of the mark M is good, each process in the next device (for example, pasting process, temporary crimping process, main crimping process) is executed without any abnormality. . On the other hand, as shown in FIGS. 6(b) to 6(d), if the pass/fail determination result is bad, each device is stopped and an operator performs confirmation.

For example, as shown in FIG. 6(b), if the shape is defective and cannot be corrected, the confirmation result that it is defective is input again by the operator. For example, if the shape is determined to be defective as shown in FIG. 6(c), but the center coordinates of the mark M can be recognized, the center coordinates are input by the operator and an input to change the judgment result to good is made. It will be done. In addition, as shown in FIG. 6(d), if it is determined that the position is defective, but the center coordinates of the mark M can be recognized, the center coordinates are input by the operator, and an input to change the judgment result to good is made. It will be done. In this way, if the detection of mark M is incorrectly determined to be defective even though it is not completely defective (for example, although the center coordinates of mark M can be recognized), the operator can make corrections. Input is made.

The correction input unit 72 is, for example, a keyboard and a mouse, and obtains correction inputs from the operator. The modification unit 74 modifies the determination result incorrectly determined as defective by the determination unit 73 to a modification result of good, based on the modification input acquired by the modification input unit 72. The image of the mark M, the pass/fail judgment result, and the correction result shown in FIGS. be done.

Next, an example will be described in which the quality of the ACF 6 is determined using the ACF 6 as an imaging target.

FIG. 7 is a diagram showing an image of the ACF 6 provided on the substrate 3 and a determination result of the quality of the ACF 6. FIG. 7 shows an example in which the ACF 6 attached to the substrate 3 has a rectangular shape.

FIG. 7A shows an example in which the shape and position of the image of the ACF 6 are within the reference value range, and the determination result of the quality of the ACF 6 is good. FIGS. 7(b) and 7(c) show an example in which the shape and position of the image of the ACF 6 are outside the reference value range, and the pass/fail determination result is poor.

As shown in FIG. 7A, if the determination result of the ACF 6 is good, the next process (for example, temporary pressure bonding process) is executed without any abnormality. On the other hand, as shown in FIGS. 7(b) and 7(c), if the pass/fail determination result is bad, the operation of each device is stopped and an operator performs confirmation.

For example, as shown in FIG. 7(b), if the shape is defective and cannot be corrected, a confirmation result indicating that it is defective is input again. For example, as shown in FIG. 7(c), it is determined that the shape and position are defective, but if the operator recognizes that there is no problem in performing the next process in this state, the determination result is changed to good. Input is made. In this way, even if the ACF 6 is not completely defective (for example, even though the ACF 6 completely covers the electrode part 4), it is incorrectly determined that the ACF 6 is defective. Correct input is made.

The correction input unit 72 obtains correction input from the operator. The modification unit 74 modifies the determination result incorrectly determined as defective by the determination unit 73 to a modification result of good, based on the modification input acquired by the modification input unit 72. The images of the ACF 6, the pass/fail judgment results, and the correction results shown in FIGS. Ru.

The learning model generation unit 75 performs machine learning using the image, the determination result, and the correction result as learning data, and generates a learning model for determining the quality of the imaging target. For example, the learning model generation unit 75 uses a neural network that has an input layer that inputs each pixel in an image and an output layer that outputs results of good and bad, and uses the judgment result or correction result as a label to identify the object to be imaged. A learning model is generated so that the correct answer rate for pass/fail judgments is high. If both the determination result and the correction result exist, learning is performed using the correction result as a label.

Generation of this learning model may be executed by a computer that controls each device of the component mounting system 1, or may be executed by a computer of an external server. Further, the generation of the learning model may be executed each time a predetermined amount of data of the images, determination results, and correction results are accumulated, or may be executed each time a correction result by the operator is input.

The learning model evaluation unit 76 determines whether the proficiency level of the learning model has reached a prescribed level. For example, the learning model evaluation unit 76 determines whether the correct answer rate based on the learning model has reached the same level as that of a skilled worker (for example, a correct answer rate of 99.9%). Note that the learning model and the prescribed level for evaluating the proficiency level are stored in the storage unit 70b.

The changing unit 77 causes the determining unit 73 to determine the quality of the imaging target using the learning model when the learning model evaluation unit 76 evaluates that the proficiency level of the learning model has reached a prescribed level. The method of determination by the determination unit 73 is changed.

For example, the changing unit 77 changes the determination method of the determining unit 73 so that instead of determining the quality of the imaging target based on the detected value and the reference value, the learning model is used to determine the quality of the imaging target. May be changed. In addition, the changing unit 77 changes the determination method of the determining unit 73 so that it performs both the quality determination of the imaging target based on the detected value and the reference value and the quality determination of the imaging target using the learning model. You may.

In this way, the computer 70 determines whether the imaged object is good or bad by using the learning model generated based on the image, the judgment result of the judgment section 73, and the correction result of the correction section 74. Misjudgment can be suppressed when making a determination. Thereby, a decrease in productivity in the component mounting system 1 can be suppressed.

The above-mentioned change in the method of determination may be executed by the operator's judgment, or by the judgment of the computer 70 based on the evaluation result of the learning model evaluation unit 76.

Further, the above-mentioned learning model is switched for each product type. When there are many production types, it takes time to learn from the beginning, so a learning model of a similar production type may be selected from the learned learning models and learning may be performed based on that learning model. This can shorten the time it takes to create a learning model.

[4. Component mounting method]
Next, a component mounting method according to an embodiment will be described. The component mounting method according to the present embodiment includes a first mode production process S1 that does not use a learning model, and a second mode production process S2 that uses a learning model. The component mounting method also includes a changing step S1A that changes the first mode production step S1 to the second mode production step S2.

FIG. 8 is a diagram illustrating an example of the steps included in the first mode production process S1. The first mode production process S1 shown in FIG. 8 includes an image acquisition process S10, a first determination process S21, a correction process S30, and an attachment process S40. In this example, the image acquisition step S10 is a step of acquiring an image of the mark M taken by the pasting section 20, and the attaching step S40 is a pasting step of pasting the ACF 6 onto the substrate 3 using the pasting section 20.

First, the image acquisition unit 71 of the computer 70 acquires an image of the mark M provided on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires an image of the mark M captured by the imaging unit 29 of the pasting unit 20.

Next, the determination unit 73 determines the quality of the mark M based on the detected value determined from the image and a first reference value set in advance (S21). Examples of the detected values include the position, area, shape, color, etc. of the mark M within the imaging range (imaging area). If the determination result of the determination unit 73 is good (passed in S21), the position of the substrate 3 is adjusted based on the image of the mark M (S29).

When the determination result of the determination unit 73 is defective (defective in S21), it is determined whether the determination result is an erroneous determination (S25). If the determination result is correct, that is, if it is not an erroneous determination (No in S25), the board 3 is removed from the component mounting system 1 (S50).

On the other hand, if the determination result is an erroneous determination (Yes in S25), the determination result is corrected to be good (S30). Specifically, the correction unit 74 corrects the determination result of the mark M as good or bad based on the correction input obtained by the correction input unit 72. The position of the substrate 3 having the mark M whose determination result has been corrected is adjusted based on the image of the mark M (S29).

Once the position of the board 3 is adjusted, the mounting section performs a component mounting process (S40). The attachment section in this example is the attachment mechanism 22, and the component attachment process is an attachment process for attaching the ACF 6 to the board 3.

The process shown in FIG. 8 is not limited to the pasting part 20, but can also be applied to the temporary crimping part 30 or the main crimping part 40. For example, when the process shown in FIG. 8 is applied to the temporary pressure bonding section 30, the step of acquiring the image of the mark M taken by the temporary pressure bonding section 30 corresponds to the image acquisition step S10, and the electronic component 5 is temporarily bonded to the board 3. This process corresponds to the attachment process S40. Furthermore, when the process shown in FIG. 8 is applied to the main crimping section 40, the step of acquiring the image of the mark M taken by the main crimping section 40 corresponds to the image acquisition step S10, and the electronic component 5 is main crimped onto the board 3. This process corresponds to the attachment process S40.

The image of the mark M, the pass/fail determination result, and the correction result obtained in the first mode production process S1 are output to the computer 70. In the computer 70, a learning process S60 is executed in parallel with the first mode production process S1.

FIG. 9 is a flowchart illustrating an example of generating a first learning model based on data of the first mode production process S1 and changing the first mode production process S1 to the second mode production process S2.

The learning process S60 includes a learning model generation process S61 that generates a first learning model, and a learning model evaluation process S62 that evaluates the first learning model.

The learning model generation unit 75 performs machine learning using the image obtained in the image acquisition step S10 of the production step S1 in the first mode, the determination result of the first determination step S21, and the modification result of the modification step S30 as learning data. A first learning model for determining the quality of the mark M is generated (S61).

The learning model evaluation unit 76 evaluates the proficiency level of the first learning model (S62). The proficiency level is determined by, for example, whether the correct answer rate of the first learning model is equal to or higher than a specified value. This correct answer rate is set to the same level as that of a skilled worker. If it is not evaluated that the proficiency level of the first learning model has reached the prescribed level (No in S62), the process returns to the learning model generation step S61 and continues to generate the first learning model. On the other hand, if it is evaluated that the proficiency level of the first learning model has reached the prescribed level (Yes in S62), the learning model evaluation unit 76 generates an arrival signal indicating that the proficiency level has reached the prescribed level. is output to the changing unit 77.

The changing unit 77 determines whether the first mode production process S1 should be changed to the second mode production process S2 (S1A). If the change unit 77 does not receive the arrival signal, it determines that the production process should not be changed (No in S1A), and continues to execute the production process S1 in the first mode. On the other hand, when the change unit 77 receives the arrival signal (Yes in S62), it determines that the production process should be changed (Yes in S1A), and changes the production process S1 in the first mode to the production process in the second mode. Change to S2.

FIG. 10 is a diagram illustrating an example of the steps included in the second mode production process S2. The second mode production process S2 shown in FIG. 10 includes an image acquisition process S10, a second determination process S22, and an attachment process S40. In this example as well, the image acquisition step S10 is a step of acquiring an image of the mark M taken by the pasting section 20, and the attaching step S40 is a pasting step of pasting the ACF 6 on the substrate 3 using the pasting section 20.

First, the image acquisition unit 71 acquires an image of the mark M provided on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires an image of the mark M captured by the imaging unit 29 of the pasting unit 20.

Next, the determining unit 73 uses the first learning model to determine the quality of the mark M (S22). If the judgment result of the judgment unit 73 is defective (defective in S22), the board 3 is removed from the component mounting system 1 (S50). If the determination result of the determination unit 73 is good (passed in S22), the position of the substrate 3 is adjusted based on the image of the mark M (S29).

Once the position of the board 3 is adjusted, the mounting section performs a component mounting process (S40). The attachment section in this example is the attachment mechanism 22, and the component attachment process is an attachment process for attaching the ACF 6 to the board 3.

The process shown in FIG. 10 is not limited to the pasting part 20, but can also be applied to the temporary pressure bonding part 30 or the main pressure bonding part 40. For example, when the process shown in FIG. 10 is applied to the temporary pressure bonding section 30, the step of acquiring the image of the mark M taken by the temporary pressure bonding section 30 corresponds to the image acquisition step S10, and the electronic component 5 is temporarily bonded to the board 3. This process corresponds to the attachment process S40. Furthermore, when the process shown in FIG. 10 is applied to the main crimping section 40, the step of acquiring the image of the mark M taken by the main crimping section 40 corresponds to the image acquisition step S10, and the electronic component 5 is main crimped onto the board 3. This process corresponds to the attachment process S40.

[5. Other examples of component mounting methods]
Next, another example of the component mounting method will be described.

FIG. 11 is a diagram showing another example of the steps included in the first mode production step S1. The first mode production process S1 shown in FIG. 11 includes an image acquisition process S10, a first determination process S21, a correction process S30, and an attachment process S40. In this example, a case will be explained in which the image acquisition step S10 is a step of acquiring an image of the ACF 6 pasted by the pasting section 20, and the attachment step S40 is a temporary pressure bonding step and a main pressure bonding step following the pasting step. do.

First, the image acquisition unit 71 acquires an image of the ACF 6 pasted on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires an image of the ACF 6 captured by the imaging unit 29 of the pasting unit 20.

Next, the determination unit 73 determines the quality of the ACF 6 based on the detected value determined from the image and a preset second reference value (S21). Examples of the detected values include the position, area, shape, color, etc. of the ACF 6 within the imaging range (imaging region). If the determination result of the determination unit 73 is good (passed in S21), the next attachment step S40 is executed. As described above, the attachment process S40 is a temporary pressure bonding process and a main pressure bonding process.

When the determination result of the determination unit 73 is defective (defective in S21), it is determined whether the determination result is an erroneous determination (S25). If the determination result is correct, that is, if it is not an erroneous determination (No in S25), the board 3 is removed from the component mounting system 1 (S50).

On the other hand, if the determination result is an erroneous determination (Yes in S25), the determination result is corrected to be good (S30). Specifically, the modification section 74 modifies the determination result of the quality of the ACF 6 to "good" based on the modification input acquired by the modification input section 72. The next mounting step S40 is executed for the board 3 having the ACF 6 for which the judgment result has been corrected.

The image of the ACF 6, the quality determination result, and the correction result obtained in the first mode production process S1 are output to the computer 70. In the computer 70, a learning process S60 is executed in parallel with the first mode production process S1.

FIG. 12 shows an example in which a second learning model is generated based on the data of the first mode production process S1 shown in FIG. 11, and the first mode production process S1 is changed to the second mode production process S2. It is a flowchart.

The learning process S60 includes a learning model generation process S61 that generates a second learning model, and a learning model evaluation process S62 that evaluates the second learning model.

The learning model generation unit 75 performs machine learning using the image obtained in the image acquisition step S10 of the production step S1 in the first mode, the determination result of the first determination step S21, and the modification result of the modification step S30 as learning data. A second learning model for determining the quality of the ACF 6 is generated (S61).

The learning model evaluation unit 76 evaluates the proficiency level of the second learning model (S62). The proficiency level is determined by, for example, whether the correct answer rate of the second learning model is equal to or higher than a specified value. This correct answer rate is set to the same level as that of a skilled worker. If it is not evaluated that the proficiency level of the second learning model has reached the prescribed level (No in S62), the process returns to the learning model generation step S61 and continues to generate the second learning model. On the other hand, when it is evaluated that the proficiency level of the second learning model has reached the prescribed level (Yes in S62), the learning model evaluation unit 76 generates an arrival signal indicating that the proficiency level has reached the prescribed level. is output to the changing unit 77.

The changing unit 77 determines whether the first mode production process S1 should be changed to the second mode production process S2 (S1A). If the change unit 77 does not receive the arrival signal, it determines that the production process should not be changed (No in S1A), and continues to execute the production process S1 in the first mode. On the other hand, when the changing unit 77 receives the arrival signal, it determines that the production process should be changed (Yes in S1A), and changes the first mode production process S1 to the second mode production process S2.

FIG. 13 is a diagram showing an example of the steps included in the second mode production step S2. The second mode production process S2 shown in FIG. 13 includes an image acquisition process S10, a second determination process S22, and an attachment process S40. In this example as well, the image acquisition step S10 is a step of acquiring an image of the ACF 6 pasted by the pasting section 20, and the attachment step S40 is a temporary pressure bonding step and a main pressure bonding step following the pasting step.

First, the image acquisition unit 71 acquires an image of the ACF 6 provided on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires an image of the ACF 6 captured by the imaging unit 29 of the pasting unit 20.

Next, the determination unit 73 determines the quality of the ACF 6 using the second learning model (S22). If the judgment result of the judgment unit 73 is defective (defective in S22), the board 3 is removed from the component mounting system 1 (S50). If the determination result of the determination unit 73 is good (passed in S22), the next attachment step S40 is executed. As described above, the attachment process S40 is a temporary crimping process and a main crimping process.

[6. Modification example 1 of embodiment]
Next, a component mounting method according to Modification 1 of the embodiment will be described. In modification example 1, the second determination step in the production process S2 of the second mode includes both the quality determination of the imaging target based on the detected value and the reference value, and the quality determination of the imaging target using a learning model. I will explain about it. In this modified example 1, the alignment mark M and the ACF 6 are each taken as an imaging target, and the ACF 6 and the electronic component 5 are each described as a component.

FIG. 14 is a flowchart showing the production process S2 in the second mode of the component mounting method according to the first modification. The production process S2 in the second mode shown in FIG. 14 includes an image acquisition process S10, a first determination process S21, a second determination process S22, and an attachment process S40.

First, the image acquisition unit 71 acquires an image of the imaging target provided on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires images of the imaging target captured by the imaging units 29, 39, and 49.

Next, the determination unit 73 determines the quality of the imaging target based on the detected value obtained from the image and a preset reference value (S21). Examples of the detected values include the position, area, shape, color, etc. of the imaging target within the imaging range (imaging area). If the determination result of the determination unit 73 is good (passed in S21), the attachment unit performs component attachment processing (S40).

If the determination result of the determination unit 73 is defective (defect in S21), the quality of the imaging target is determined using the learning model generated based on the data of the first mode production process S1 (S22). If the result of the quality determination using the learning model is negative (defective in S22), the board 3 is removed from the component mounting system 1 (S50).

On the other hand, if the determination result using the learning model is good (passed in S22), a component mounting process is performed by the mounting section (S40). Note that the component mounting process by the mounting unit is a pasting process in which the ACF 6 is pasted on the board 3 by the pasting mechanism 22, a temporary pressure bonding process in which the electronic component 5 is temporarily pressed on the board 3 by the component mounting mechanism 32, or a temporary pressure bonding process in which the electronic component 5 is temporarily pressed on the board 3 by the component mounting mechanism 32, or This is a final crimping process in which the electronic component 5 is fully crimped onto the substrate 3.

[7. Modification 2 of embodiment]
Next, a component mounting method according to a second modification of the embodiment will be described. In modification 2, an example will be described in which it is determined whether or not to perform the second determination step S22 based on the magnitude of the difference between the detected value and the reference value. In the second modification as well, the alignment mark M and the ACF 6 are each taken as an imaging target, and the ACF 6 and the electronic component 5 are each described as a component.

FIG. 15 is a flowchart showing the second mode production process S2 of the component mounting method of Modification 2. The production process S2 in the second mode shown in FIG. 15 includes an image acquisition process S10, a first determination process S21, a second determination process S22, and an attachment process S40, and further includes a determination process S26 shown below. .

First, the image acquisition unit 71 acquires an image of the imaging target provided on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires an image of the imaging target captured by the imaging unit 29, 39, or 49.

Next, the determination unit 73 determines the quality of the imaging target based on the detected value obtained from the image and a preset reference value (S21). If the determination result of the determination unit 73 is good (passed in S21), the attachment unit performs component attachment processing (S40).

If the determination result of the determination unit 73 is negative (defective in S21), the computer 70 uses the learning model to determine the value based on the magnitude of the difference between the detected value obtained from the image and the preset reference value. It is determined whether or not the quality of the imaging target should be determined (S26). If the difference between the detected value and the reference value is larger than a predetermined value (Yes in S26), the board 3 is removed from the component mounting system 1 without making a pass/fail determination using the learning model. (S50).

On the other hand, if the difference between the detected value and the reference value is smaller than a predetermined value (No in S26), a quality determination is performed in a second determination step using a learning model (S22). The subsequent steps are similar to the component mounting method according to the first modification.

[8. Modification example 3 of embodiment]
Next, a component mounting method according to modification 3 of the embodiment will be described. In modification 3, an example will be described in which a learning model is generated in the production process S2 of the second mode. Also in the third modification, the alignment mark M and the ACF 6 will be described as imaging targets, and the ACF 6 and the electronic component 5 will be described as a component.

FIG. 16 is a flowchart illustrating an example of generating a learning model in the second mode production process S2.

The production process S2 in the second mode shown in FIG. 16 includes an image acquisition process S10, a second determination process S22, and an attachment process S40, and further includes a correction process S31 shown below.

First, the image acquisition unit 71 acquires an image of the imaging target provided on the substrate 3 (S10). Specifically, the image acquisition unit 71 acquires an image of the imaging target captured by the imaging unit 29, 39, or 49.

Next, the determination unit 73 determines the quality of the imaging target using the learning model of the production process S2 in the second mode (S22). When the determination result of the determination unit 73 is good (passed in S22), the next attachment step S40 is executed.

If the determination result of the determination unit 73 is defective (defective in S22), it is determined whether the determination result is an erroneous determination (S25A). If the determination result is correct, that is, if it is not an erroneous determination (No in S25A), the board 3 is removed from the component mounting system 1 (S50).

On the other hand, if the determination result is an erroneous determination (Yes in S25A), the determination result is corrected to be good (S31). Specifically, the modification unit 74 modifies the determination result of the quality of the imaging target to good based on the modification input acquired by the modification input unit 72. The next mounting step S40 is performed on the board 3 having the imaging target for which the determination result has been corrected.

The image of the imaging target acquired in the production process S2 of the second mode, the quality determination result, and the correction result are output to the computer 70. In the computer 70, a learning process (S60) is executed in parallel with the second mode production process S2 (see FIG. 9).

Similarly to FIG. 9, the learning model generation unit 75 learns the image obtained in the image acquisition step S10 of the production step S2 in the second mode, the determination result of the second determination step S22, and the modification result of the modification step S31. Machine learning is performed on the data to generate a learning model for determining the quality of the imaging target (S61).

The learning model evaluation unit 76 evaluates the proficiency level of the learning model (S62). The proficiency level is determined, for example, by whether the correct answer rate of the learning model is equal to or higher than a specified value. The correct answer rate of the learning model is set to a higher correct answer rate than the correct answer rate of the original learning model. If it is evaluated that the proficiency level of the learning model has reached the prescribed level (Yes in S62), the computer 70 changes the original learning model to the learning model and executes the second mode production process S2. do.

(summary)
The component mounting method according to the present embodiment includes a first mode production process S1, a learning model generation process, a learning model evaluation process, and a second mode production process S2. and a changing step of changing to.

The production process S1 in the first mode includes an image acquisition step of acquiring an image of the imaging target provided on the substrate 3, and a determination of the quality of the imaging target based on a detected value obtained from the image and a preset reference value. a first judgment step to be carried out, a correction step of correcting the judgment result of the first judgment step to good when it is incorrectly judged as defective in the first judgment step, and attaching the component to the board 3 having the imaging target judged as good. Has an installation process. The learning model generation process is for determining the quality of the imaging target by machine learning using the image acquired in the production process S1 of the first mode, the determination result of the first determination process, and the modification result of the modification process as learning data. Generate a learning model for. The learning model evaluation step evaluates the proficiency level of the learning model generated in the learning model generation step. In the change step, when it is evaluated that the proficiency level of the learning model has reached a specified level in the learning model evaluation step, the production process S1 of the first mode is changed to the first mode in which the quality of the imaging target is determined using the learning model. The production process is changed to a second mode production process S2 having two determination processes.

In this way, the production process S1 of the first mode, which determines the quality of the imaging target based on the detection value obtained from the image and the preset reference value, is performed based on the data of the production process S1 of the first mode. By changing to the second mode production process S2, which includes a step of performing quality determination using the generated learning model, it is possible to suppress erroneous determination when determining quality of the imaging target. Thereby, it is possible to suppress a decrease in productivity in component mounting (component attachment processing).

For example, in the production process S1 of the first mode, the reference value may be set more strictly in order to prevent defects from occurring frequently. However, even in such a case, the production process S1 of the first mode can be changed to the production process S2 of the second mode, which performs a quality judgment using a learning model generated based on the data of the production process S1 of the first mode. By changing, it is possible to suppress erroneous determination when determining the quality of the imaging target.

Moreover, the changing process may change the production process S1 in the first mode to the production process S2 in the second mode by replacing the first determination process with a second determination process.

In this way, by changing the first mode production process S1 to the second mode production process S2 using a learning model, it is possible to suppress erroneous determinations when determining the quality of the imaged object. Thereby, it is possible to suppress a decrease in productivity in component mounting.

The production process S2 in the second mode includes an image acquisition process, a first determination process, and a second determination process, and when the imaging target is determined to be defective in the first determination process, the second determination process is executed. You may.

In this way, by executing the second determination step when the object is determined to be defective in the first determination step, it is possible to suppress erroneous determination when determining the quality of the imaging target. Thereby, it is possible to suppress a decrease in productivity in component mounting.

In addition, in the production process S2 of the second mode, when it is determined to be defective in the first determination process, it is determined whether or not to perform the second determination process based on the magnitude of the difference between the detected value and the reference value. It may further include a determination step.

In this way, by further including the determination step of determining whether or not to perform the second determination step, it is possible to reduce the load applied when performing the pass/fail determination using the second step. Thereby, it is possible to suppress a decrease in productivity in component mounting.

Further, the learning model generation process and the learning model evaluation process may be repeatedly executed in the first mode production process S1.

According to this, it is possible to improve the proficiency level of the learning model, and therefore it is possible to suppress erroneous determination when determining the quality of the imaging target. Thereby, it is possible to suppress a decrease in productivity in component mounting.

In addition, the learning model generation step further includes the image, the determination result of the second determination step, and a step of correcting the determination result of the second determination step to good when it is incorrectly determined as defective in the second determination step. A learning model for determining the quality of the imaging target may be generated by machine learning using the correction results as learning data, and the second determination step may use the learning model to determine the quality of the imaging target.

According to this, the proficiency level of the learning model can be further improved, so that erroneous determinations can be suppressed when determining the quality of the imaging target. Thereby, it is possible to suppress a decrease in productivity in component mounting.

Further, the imaging target may be the alignment mark M provided on the board 3, and the component to be subjected to the mounting process may be the ACF 6 or the electronic component 5.

According to this, it is possible to suppress erroneous determination when determining the quality of the alignment mark M. Thereby, it is possible to suppress a decrease in productivity in component mounting.

Further, the imaging target may be the ACF 6 provided on the board 3, and the component in the mounting process may be the electronic component 5.

According to this, it is possible to suppress erroneous determination when determining the quality of the ACF 6. Thereby, it is possible to suppress a decrease in productivity in component mounting.

Further, in the learning model generation step, machine learning may be started based on a learned model selected from a plurality of learned models prepared in advance according to the type of board 3.

According to this, the time to generate a learning model can be shortened. Thereby, it is possible to suppress a decrease in productivity in component mounting.

The component mounting system 1 according to the present embodiment includes an image acquisition unit 71 that acquires an image of an imaging target provided on the board 3, and an image acquisition unit 71 that acquires an image of an imaging target provided on the board 3, and an image acquisition unit 71 that acquires an image of an object to be imaged. A determining section 73 that determines whether the object is good or bad; a modifying section 74 that modifies the determination result of the determining section 73 to good when the determining section 73 erroneously determines that the object is defective; and a mounting section for mounting components onto a substrate 3 having an imaged object. The component mounting system 1 also uses machine learning that uses the image, the judgment result of the judgment unit 73, and the correction result of the correction unit 74 as learning data to generate a learning model for determining the quality of the imaged object. It further includes a model generation section 75 and a learning model evaluation section 76 that evaluates the proficiency level of the learning model generated by the learning model generation section 75. When the learning model evaluation unit 76 evaluates that the proficiency level of the learning model has reached a prescribed level, the determining unit 73 uses the learning model to determine the quality of the imaging target.

In this way, the determination unit 73 determines the quality of the imaging target by using the learning model generated based on the image, the determination result of the determination unit 73, and the modification result of the modification unit 74. It is possible to suppress erroneous judgments when making pass/fail judgments. Thereby, a decrease in productivity in the component mounting system 1 can be suppressed. Note that the above attachment section is, for example, the pasting mechanism 22, the component mounting mechanism 32, or the crimping mechanism 42.

(Other embodiments)
Although the component mounting system and the like according to the present embodiment have been described above based on the above embodiments, the present invention is not limited to the above embodiments.

For example, in the embodiment described above, an example is shown in which the learning model is used in determining the quality of the mark M in the pasting process, temporary pressure bonding process, and main pressure bonding process, and in determining the quality of the ACF6 in the pasting process. The quality determination and the quality determination of the ACF 6 do not need to be performed in all of the above steps, and may be performed in at least one step.

In addition, in the above embodiment, machine learning is performed using the images, determination results, and correction results obtained in the pasting process, preliminary crimping process, and main crimping process as learning data, but the machine learning is not limited to this, and Machine learning may also be performed by including inspection results obtained in the process in the learning data. For example, by assigning IDs to individual production objects (for example, display panels), it becomes possible to reflect inspection results in subsequent processes in the learning model. According to this, when the cause of a defective inspection in the subsequent process is in the pasting process, temporary crimping process, and main crimping process, the learning model can be corrected based on the inspection results.

For example, in the above embodiments, all or part of the computer components may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. . Each component may be realized by a program execution unit such as a CPU (Central Processing Unit) or a processor reading and executing a software program recorded on a recording medium such as an HDD (Hard Disk Drive) or a semiconductor memory. good.

Also, the components of the computer may be comprised of one or more electronic circuits. Each of the one or more electronic circuits may be a general-purpose circuit or a dedicated circuit.

The one or more electronic circuits may include, for example, a semiconductor device, an IC (Integrated Circuit), an LSI (Large Scale Integration), or the like. An IC or LSI may be integrated into one chip or into multiple chips. Here, it is called an IC or LSI, but the name changes depending on the degree of integration, and may be called a system LSI, VLSI (Very Large Scale Integration), or ULSI (Ultra Large Scale Integration). Furthermore, an FPGA (Field Programmable Gate Array) that is programmed after the LSI is manufactured can also be used for the same purpose.

In addition, forms obtained by applying various modifications to each embodiment that those skilled in the art can think of, or by arbitrarily combining the constituent elements and functions of each embodiment without departing from the spirit of the present invention. The form is also included in the present invention.

INDUSTRIAL APPLICATION This invention can be utilized for the component mounting system etc. which produce a liquid crystal panel.

1 Component mounting system 1a, 1b, 1c base 3 board 4 electrode part 5 electronic component 6 ACF (anisotropic conductive member)
6a ACF tape 10 Board loading section 11, 23, 33, 43, 51 Stage 20 Pasting section 21, 31, 41 Stage moving section 22 Pasting mechanism 24 Beam 25 Pasting head 25a Tape supply section 25b Pasting tool 26 Pasting support base 29, 39 , 49, imaging section 30 temporary crimping section 32 component mounting mechanism 34 component supply section 40 main crimping section 42 crimping mechanism 44 crimping support section 45 crimping unit 47 pressing mechanism 48 crimping head 50 substrate unloading section 60 transport section 61 moving base 62A, 62B, 62C, 62D Substrate transport mechanism 63 Base 64 Arm unit 70 Computer 70a Control section 70b Storage section 71 Image acquisition section 72 Correction input section 73 Judgment section 74 Correction section 75 Learning model generation section 76 Learning model evaluation section 77 Changing section 79 Display Part M Alignment mark (mark)

Claims (10)

an image acquisition step of acquiring an image of an imaging target provided on a substrate; a first determination step of determining the quality of the imaging target based on a detected value obtained from the image and a preset reference value; A first mode including a correction step of correcting the determination result of the first determination step to good when the determination result is incorrectly determined as defective in the determination step, and a mounting step of attaching the component to a board having the imaging target determined to be good. production process and
Learning for determining the quality of the imaging target by machine learning using the image acquired in the production process of the first mode, the determination result of the first determination process, and the modification result of the modification process as learning data. a learning model generation step for generating a model;
a learning model evaluation step of evaluating the proficiency level of the learning model generated in the learning model generation step;
When it is evaluated that the proficiency level of the learning model has reached a prescribed level in the learning model evaluation step, the production process of the first mode is replaced with a step of determining the quality of the imaging target using the learning model. a change process of changing to a second mode production process having two determination processes;
Component mounting methods including. The component mounting method according to claim 1, wherein the changing step changes the first mode production process to the second mode production process by replacing the first judgment process with the second judgment process. The production process in the second mode includes the image acquisition process, the first determination process, and the second determination process, and when the imaging target is determined to be defective in the first determination process, the production process in the second mode includes the image acquisition process, the first determination process, and the second determination process. The component mounting method according to claim 1, further comprising performing a determination step. The production process in the second mode determines whether or not to perform the second determination process based on the magnitude of the difference between the detected value and the reference value when it is determined to be defective in the first determination process. The component mounting method according to claim 3, further comprising a determination step of determining. The component mounting method according to any one of claims 1 to 4, wherein the learning model generation step and the learning model evaluation step are repeatedly executed in the first mode production process. The learning model generation step further corrects the image, the determination result of the second determination step, and the determination result of the second determination step to be good if the second determination step incorrectly determines that the image is defective. Generate a learning model for determining the quality of the imaging target by machine learning using the process correction results as learning data,
6. The component mounting method according to claim 1, wherein the second determination step uses the learning model to determine the quality of the imaging target. The imaging target is an alignment mark provided on the substrate,
7. The component mounting method according to claim 1, wherein the component in the mounting step is an ACF (Anisotropic Conductive Film) or an electronic component. The imaging target is an ACF provided on the substrate,
The component mounting method according to any one of claims 1 to 6, wherein the component in the mounting step is an electronic component. According to any one of claims 1 to 8, in the learning model generation step, the machine learning is started based on a trained model selected from a plurality of trained models prepared in advance according to the type of the board. The described component mounting method. an image acquisition unit that acquires an image of an imaging target provided on the board;
a determination unit that determines the quality of the imaging target based on a detected value obtained from the image and a preset reference value;
a correction unit that corrects the determination result of the determination unit to good when the determination unit erroneously determines that the determination unit is defective;
a mounting unit that attaches a component to a board having the imaging target determined to be good by the determination unit or the correction unit;
A component mounting system comprising:
a learning model generation unit that generates a learning model for determining the quality of the imaging target by machine learning using the image, the determination result of the determination unit, and the modification result of the modification unit as learning data;
a learning model evaluation unit that evaluates the proficiency level of the learning model generated by the learning model generation unit;
Furthermore,
The determination unit determines the quality of the imaging target using the learning model when the learning model evaluation unit evaluates that the proficiency level of the learning model has reached a prescribed level.

Images