KR20200006028A - Printed circuit board inspecting apparatus, method for determining fault type of screen printer and computer readable recording medium - Google Patents

Abstract

Disclosed is a printed circuit board inspecting apparatus. The printed circuit board inspecting apparatus comprises: a memory to store a machine learning-based model learned to derive at least one value indicating the relation of one or more fault types of a screen printer to an image indicating whether a plurality of solder pastes printed on a printed circuit board are abnormal; and a processor to generate at least one image indicating whether a plurality of solder pastes printed on a first printed circuit board are abnormal if an abnormality of at least one solder paste among the plurality of solder pastes is sensed by using an image for the first printed circuit board, use the machine learning-based model to acquire at least one value indicating the relation of the one or more one fault types to the generated at least one image, and determine a fault type associated with the sensed at least one solder paste among the one or more fault types based on the acquired at least one value.

Description

PRINTED CIRCUIT BOARD INSPECTING APPARATUS, METHOD FOR DETERMINING FAULT TYPE OF SCREEN PRINTER AND COMPUTER READABLE RECORDING MEDIUM}

TECHNICAL FIELD The present disclosure relates to a printed circuit board inspection apparatus, and more particularly, to a printed circuit board inspection apparatus for determining a defect type of a screen printer used for printing a printed circuit board.

The present disclosure is derived from a study conducted as a part of the robot industry convergence core technology development project of the Ministry of Trade, Industry and Energy [Task No .: 10077589, Project title: Development of SMT optimization system technology based on machine learning].

A process of printing solder paste on a printed circuit board mounted on an electronic device is performed by a screen printer, and a process of printing solder paste on a printed circuit board by a screen printer is as follows. The screen printer places the printed circuit board on a table for fixing the printed circuit board, and aligns the stencil on the printed circuit board so that the opening of the stencil is located on the pad of the corresponding printed circuit board. The screen printer then uses a squeegee to print the solder paste onto the printed circuit board through the openings in the stencil. The screen printer then separates the stencil and the printed circuit board.

The shape of the solder paste printed on the printed circuit board may be inspected through solder paste inspection (SPI) technology. SPI technology is a technique of obtaining a two-dimensional or three-dimensional image of the solder paste printed on the printed circuit board through the optical technology, and to examine the shape of the solder paste printed on the printed circuit board from the obtained image.

The present disclosure provides a printed circuit board inspection apparatus capable of determining a fault type of a screen printer associated with anomaly of solder paste using a machine-learning based model.

The present disclosure discloses a computer-readable recording medium having recorded thereon a program comprising executable instructions for causing a printed circuit board inspection apparatus to use a machine-learning based model to determine a defect type of a screen printer associated with an abnormality of solder paste. To provide.

The present disclosure provides a method for determining a defect type of a screen printer related to an abnormality of solder paste using a machine-learning based model in a printed circuit board inspection apparatus.

According to one embodiment of the present disclosure, an apparatus for inspecting a printed circuit board includes at least one indicating a relationship between at least one defect type of a screen printer and an image indicating whether a plurality of solder pastes printed on the printed circuit board are abnormal. It may include a memory for storing the machine-learning based model trained to derive a value of and a processor electrically connected with the memory. When an abnormality of at least one solder paste of a plurality of solder pastes printed on the first printed circuit board is detected by using an image of the first printed circuit board, the processor determines whether the plurality of solder pastes are abnormal. Generate at least one image representative, and use the machine-learning based model to obtain at least one value indicating a relationship between the at least one defect type and the generated at least one image, wherein the acquired at least Based on one value, a defect type associated with at least one solder paste in which the abnormality is detected among the at least one defect type may be determined.

In one embodiment, the processor uses the first measurement information for the plurality of solder pastes obtained through an image of the first printed circuit board, the at least one solder of the plurality of solder pastes. It can detect abnormality of paste.

In an embodiment, the first measurement information may include at least one of a volume, an area, a height, a shape, or a position of the solder paste.

In example embodiments, the memory may include a plurality of solder pastes printed on each of the plurality of second printed circuit boards, which have been checked for abnormality by the printed circuit board inspection apparatus before the first printed circuit board. The apparatus further stores the second measurement information, and the processor may detect whether the at least one solder paste is abnormal among the plurality of solder pastes by using the first measurement information and the second measurement information.

In example embodiments, the second measurement information may include measurement information regarding a plurality of solder pastes printed on each of the plurality of second printed circuit boards according to an inspection order of each of the plurality of second printed circuit boards. Can be generated by combining time series.

In an embodiment, the processor may further include a plurality of solder pastes printed on each of the plurality of second printed circuit boards according to an inspection order of each of the plurality of second printed circuit boards, based on the second measurement information. At least one solder paste printed on the first printed circuit board of the at least one defect type is determined based on a change of the measurement information on the determined plurality of solder pastes. The type of defect associated with the abnormality can be determined.

In one embodiment, the at least one defect type is a squeeze blade defect, a squeeze blade fixation defect, a support defect, a table defect, a defect due to poor grid lock setting, a defect due to bad solder paste and a poor paste paste, And at least one of a defect due to a lack of solder paste and a defect due to a poor stencil contact.

In one embodiment, the machine-learning based model may be a convolution neural network (CNN) model.

In one embodiment, the processor may generate the machine-learning based model through a plurality of images indicating whether a plurality of solder pastes printed on each of the plurality of printed circuit boards classified according to the at least one defect type are abnormal. I can learn.

In an embodiment, the processor may determine a defect type associated with at least one solder paste in which the abnormality is detected, and then train the machine-learning based model through the generated at least one image.

In an embodiment, the processor may determine a defect type corresponding to at least one value of the obtained at least one value that is greater than or equal to a preset threshold value as a defect type associated with the at least one solder paste in which the abnormality is detected. have.

According to one embodiment of the present disclosure, in a non-transitory computer readable recording medium having recorded thereon a program for performing on a computer, the program may, when executed by a processor, cause the processor to display an image of a first printed circuit board. If an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board is detected, generating at least one image indicating whether the plurality of solder pastes are abnormal, a screen printer The at least one defect using a machine-learning based model trained to derive at least one value representative of at least one defect type and an image indicating whether the plurality of solder pastes printed on the printed circuit board are abnormal; Type and the association of the generated at least one image Acquiring at least one value representing a property and determining a defect type associated with at least one solder paste in which the abnormality of the at least one defect type is detected based on the obtained at least one value. It can contain executable instructions to make them work.

In one embodiment, the executable command is the at least one of the plurality of solder pastes using first measurement information for the plurality of solder pastes obtained through an image of the first printed circuit board. Detecting whether the solder paste is abnormal may be further performed.

In an embodiment, the first measurement information may include at least one of a volume, an area, a height, a shape, or a position of the solder paste.

In an embodiment, detecting whether the at least one solder paste is abnormal among the plurality of solder pastes may include detecting at least one of the at least one solder paste using the first measurement information and the second measurement information. And detecting the abnormality of one solder paste, wherein the second measurement information comprises: a plurality of second printed circuits which are inspected for abnormality by the printed circuit board inspection apparatus before the first printed circuit board; According to an inspection order of each of the substrates, measurement information about a plurality of solder pastes printed on each of the plurality of second printed circuit boards may be combined in time series to be generated.

In one embodiment, the machine-learning based model may be a convolution neural network (CNN) model.

In one embodiment, the executable instruction is further configured to perform the step of training the machine-learning based model through a plurality of images indicating whether the plurality of solder pastes are classified according to the at least one defect type. Can be.

In an embodiment, the determining of the defect type may include determining a defect type associated with the detected at least one solder paste, the defect type corresponding to at least one value of the obtained at least one value that is greater than or equal to a preset threshold. It may include the step of determining.

According to an embodiment of the present disclosure, a method of determining a defect type of a screen printer in a printed circuit board inspection apparatus includes a plurality of solder pastes printed on the first printed circuit board using an image of the first printed circuit board. If at least one of the solder pastes is detected, generating at least one image indicating whether the plurality of solder pastes are abnormal, at least one defect type of a screen printer and a plurality of solders printed on a printed circuit board Using at least one value representing a relationship between the at least one defect type and the generated at least one image, using a machine-learning based model trained to derive at least one value representing a relationship with the image indicating an abnormality of pastes. Obtaining one value and the obtained at least one value And based on the at least one defect type, determining a defect type associated with at least one solder paste in which the abnormality is detected.

According to various embodiments of the present disclosure, an apparatus for inspecting a printed circuit board may include solder, in which an abnormality is detected among at least one defect type of a screen printer, by using an image indicating whether a plurality of solder pastes are printed on a printed circuit board. The type of defect associated with the paste can be determined. This allows you to quickly identify and correct the type of screen printer defect that caused the solder paste to fail.

1 is a block diagram of an apparatus for inspecting a printed circuit board according to various embodiments of the present disclosure.
2 is a flowchart of a method of determining a defect type associated with a solder paste in which an abnormality is detected, according to various embodiments of the present disclosure.
3 is a flowchart of a method of detecting an abnormality of at least one solder paste printed on a printed circuit board according to various embodiments of the present disclosure.
4A-4C illustrate graphs representing first measurement information and second measurement information, according to various embodiments of the present disclosure.
5 is a flowchart of a method of determining a defect type of a screen printer associated with at least one solder paste in which an anomaly is detected, according to various embodiments of the present disclosure.
6 is a graph illustrating a plurality of pieces of information decomposed from the second measurement information and the second measurement information according to various embodiments of the present disclosure.
7A and 7B illustrate images indicating whether a plurality of solder pastes printed on a printed circuit board are abnormal according to various embodiments of the present disclosure.
8 is a conceptual diagram for a machine learning-based model, in accordance with various embodiments of the present disclosure.

Embodiments of the present disclosure are illustrated for the purpose of describing the technical spirit of the present disclosure. The scope of the present disclosure is not limited to the embodiments set forth below or the detailed description of these embodiments.

All technical and scientific terms used in the present disclosure, unless defined otherwise, have the meanings that are commonly understood by one of ordinary skill in the art to which this disclosure belongs. All terms used in the present disclosure are selected for the purpose of more clearly describing the present disclosure, and are not selected to limit the scope of the rights according to the present disclosure.

As used in this disclosure, expressions such as "comprising", "including", "having", and the like, are open terms that imply the possibility of including other embodiments unless otherwise stated in the phrase or sentence in which the expression is included. It should be understood as (open-ended terms).

As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

Expressions such as “first”, “second”, and the like used in the present disclosure are used to distinguish a plurality of components from each other, and do not limit the order or importance of the components.

As used in this disclosure, the expression “based on” is used to describe one or more factors that affect the behavior or behavior of a decision, judgment, described in a phrase or sentence that includes the expression, which expression Does not exclude additional factors that affect decisions, actions of behaviour, or actions.

In the present disclosure, when a component is referred to as being "connected" or "connected" to another component, the component may be directly connected to or connected to the other component, or new It is to be understood that the connection may be made or may be connected via other components.

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. In the accompanying drawings, the same or corresponding components are given the same reference numerals. In addition, in the following description of the embodiments, it may be omitted to duplicate the same or corresponding components. However, even if the description of the component is omitted, it is not intended that such component is not included in any embodiment.

1 is a block diagram of an apparatus for inspecting a printed circuit board according to various embodiments of the present disclosure.

According to various embodiments of the present disclosure, the printed circuit board inspection apparatus 100 may include an image sensor 110, a memory 120, and a processor 130. Also, the printed circuit board inspection apparatus 100 may include a display 140 for outputting a result processed by the processor 130, another output device (not shown), or a communication circuit for transmitting the processed result to another electronic device. It may further include (not shown). The image sensor 110, the memory 120, the processor 130, and the display 140 may be electrically connected to transmit and receive signals.

In one embodiment, the image sensor 110 may obtain an image of a printed circuit board on which a plurality of solder pastes are printed by the screen printer 101. The image obtained by the image sensor 110 may be a two-dimensional or three-dimensional image of the printed circuit board. The image for the printed circuit board obtained by the image sensor 110 can be used to measure information for a plurality of solder pastes printed on the printed circuit board. For example, using an image of a printed circuit board, information including at least one of volume, area, height, shape or position for each of the plurality of solder pastes printed on the printed circuit board may be measured. However, this is only for the purpose of explanation and the present invention is not limited thereto. Various information that may represent characteristics of each of the plurality of solder pastes may be measured through an image of a printed circuit board. From below. 'Measurement information' may be defined as information on a plurality of solder pastes measured using an image of a printed circuit board.

In one embodiment, memory 120 may store a machine-learning based model. The machine-learning based model may be trained to derive at least one value representing the relationship between at least one defect type of the screen printer 101 and an image indicating whether the plurality of solder pastes printed on the printed circuit board are abnormal. have. The machine-learning based model may be, for example, the machine learning based model may be a convolution neural network (CNN) model. In addition, the machine-learning based model may be stored in a memory of an electronic device (eg, an external server) that is connected to the printed circuit board inspection apparatus 100 by wire or wirelessly. In this case, the printed circuit board inspection apparatus 100 may transmit / receive information for determining a defect type associated with at least one solder paste in which an abnormality is detected with an electronic device interworked with a wire or wirelessly.

In one embodiment, at least one value indicating a relationship between at least one defect type and an image indicating whether a plurality of solder pastes printed on a printed circuit board is abnormal may be a probability value. For example, the at least one value indicates that the image indicating whether the plurality of solder pastes are abnormal is a% related to the first defect type of the screen printer 101, and the image is related to the second defect type of the screen printer 101. Relevance can be indicated as b%. However, this is only for the purpose of description and the present invention is not limited thereto. Various values that may indicate relevance may be used.

In one embodiment, the at least one defect type of the screen printer 101 is at least one defect type that may result in an abnormality of the solder paste printed on the printed circuit board, among the various defect types that may occur in the screen printer 101. Can be. For example, at least one defect type of the screen printer 101 may be a squeeze blade defect of the screen printer 101, a squeeze blade fixing defect (e.g., a defect of a screw fixing to fix the squeeze blade, etc.), a screen printer 101 ), A table defect of the screen printer 101, a grid lock as a support, a defect due to a poor grid lock setting, a state of the solder paste supplied from the screen printer 101, for example, a solder paste. It may include at least one of neglect, defects due to poor solder paste, defects due to lack of solder paste supplied from the screen printer 101, and defects due to poor stencil contact.

However, this is only for the purpose of description and the present invention is not limited thereto, and various defect types of the screen printer 101 which may cause an abnormality of the solder paste printed on the printed circuit board include at least one defect type of the screen printer 101. Can be included.

In one embodiment, the machine-learning based model stored in the memory 120 is an abnormality of a plurality of solder pastes printed on each of the plurality of printed circuit boards classified according to at least one defect type of the screen printer 101. It can be learned through a plurality of images representing. For example, the machine-learning based model may include a plurality of first images indicating whether a plurality of solder pastes classified as a first defect type of the screen printer 101 are abnormal, a second defect type of the screen printer 101. It may be learned through a plurality of second images indicating whether or not a plurality of solder pastes classified as are abnormal. The machine-learning based model thus learned may output a value indicating a relationship between the input image and at least one of the first defect type and the second defect type when an image indicating whether a plurality of solder pastes are abnormal. have. A detailed learning method of the machine-learning based model will be described later.

In one embodiment, the processor 130 uses the image of the first printed circuit board obtained by the image sensor 110 to at least one solder paste of the plurality of solder pastes printed on the first printed circuit board. Can detect abnormality of. The first printed circuit board may be a printed circuit board that is an inspection object transferred to the printed circuit board inspection apparatus 100 after a plurality of solder pastes are printed in the screen printer 101. For example, the processor 130 may generate first measurement information for the plurality of solder pastes by using an image of the first printed circuit board. The processor 130 may detect whether at least one solder paste is abnormal among the plurality of solder pastes by using the generated first measurement information.

In one embodiment, the memory 120 may store second measurement information for the plurality of solder pastes printed on each of the plurality of second printed circuit boards that have been inspected before the first printed circuit board. The processor 130 may store the second measurement information generated in the inspection process of each of the plurality of second printed circuit boards in the memory 120. For example, the second measurement information may be used to detect an abnormality of a plurality of solder pastes printed on the first printed circuit board after the inspection is performed after the plurality of second printed circuit boards. . For example, the second measurement information may be coupled in time series with measurement information for a plurality of solder pastes printed on each of the plurality of second printed circuit boards according to an inspection order of each of the plurality of second printed circuit boards. Can be generated. A detailed method of generating the second measurement information will be described later.

In an embodiment, the processor 130 may detect an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board using the first measurement information and the second measurement information. have. For example, the processor 130 compares the first measurement information with the second measurement information to determine a difference between the measurement information and the second measurement information, and when the determined difference is out of a preset range, the first printing. It may be detected that there is an abnormality in at least one solder paste out of a predetermined range among the plurality of solder pastes printed on the circuit board. In addition, the processor 130 determines a difference between the first measurement information and the second measurement information, and when the determined difference does not deviate from a preset range, the processor 130 may have abnormalities in the plurality of solder pastes printed on the first printed circuit board. It can be detected as missing. However, this is only for the purpose of explanation and the present invention is not limited thereto. The processor 130 may variously analyze the first measurement information and the second measurement information statistically or time-series according to the inspection order of the printed circuit board. By using the first measurement information and the second measurement information as a method, it is possible to detect the abnormality of the at least one solder paste.

In one embodiment, the processor 130 is based on the second measurement information to the plurality of solder pastes printed on each of the plurality of second printed circuit boards according to the inspection order of each of the plurality of second printed circuit boards. The change in the measurement information can be determined. For example, the processor 130 may determine a difference in measurement information for a plurality of solder pastes printed on each of the plurality of second printed circuit boards, and the determined difference may be determined in each of the plurality of second printed circuit boards. You can determine how the test sequence is changing.

In addition, the processor 130 may determine a defect type associated with at least one solder paste in which an abnormality is detected among at least one defect type of the screen printer 101 based on a change in measurement information about the plurality of determined solder pastes. Can be determined. For example, if an abnormality of at least one solder paste is detected among the plurality of solder pastes printed on the first printed circuit board, the processor 130 is based on a change in measurement information about the determined plurality of solder pastes. Thus, the defect type associated with the abnormality of the at least one solder paste in which the abnormality is detected can be determined.

In an embodiment, when an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board is detected, the processor 130 generates at least one image indicating whether the plurality of solder pastes are abnormal. can do. The processor 130 uses at least one image indicating whether the plurality of solder pastes are abnormal in order to determine a defect type of the screen printer 101 associated with the at least one solder paste in which the abnormality is detected using a machine-learning based model. Can be generated.

For example, the at least one image indicating whether the plurality of solder pastes are abnormal may include an image indicating whether the plurality of solder pastes are abnormally printed on the first printed circuit board, and the plurality of solder pastes printed on the first printed circuit board. And at least one of an image indicating whether or not over soldering is performed on the solder pastes and an image indicating whether or not soldering is abnormally performed on the plurality of solder pastes printed on the printed circuit board.

For example, the processor 130 may generate an image indicating whether the plurality of solder pastes are printed on the first printed circuit board, or may generate abnormalities for the plurality of solder pastes printed on the first printed circuit board. An image indicating whether or not and a plurality of solder pastes printed on the PCB may be generated. However, this is only for the purpose of explanation and the present invention is not limited thereto. The processor 130 may generate an image used for learning the machine-learning based model.

In one embodiment, processor 130 uses a machine-learning based model stored in memory 120 to indicate at least one association of at least one defect type of screen printer 101 with at least one image generated. Can be obtained. The processor 130 may determine a defect type associated with at least one solder paste in which an abnormality is detected, based on the obtained values. The processor 130 may output information on the determined defect type through the display 140 or another output device (not shown). In addition, the processor 130 may transmit information about the determined defect type to another electronic device through a communication circuit (not shown) included in the printed circuit board inspection apparatus 100. In addition, the defect type to be determined may be plural.

For example, the processor 130 may determine a defect type corresponding to a value that is equal to or greater than a preset threshold among the acquired values as a defect type associated with at least one solder paste in which an abnormality is detected.

In one embodiment, the processor 130 may determine the defect type of the screen printer 101 associated with the at least one solder paste in which an anomaly is detected, and then train the machine-learning based model through the generated at least one image. have. As such, the processor 130 learns a machine-learning based model each time an inspection of the printed circuit board is performed to more accurately identify at least one defect type of the screen printer 101 associated with the solder paste in which an anomaly was detected. You can decide.

2 is a flowchart of a method of determining a defect type associated with a solder paste in which an abnormality is detected, according to various embodiments of the present disclosure.

Although process steps, method steps, algorithms, and the like have been described in time series in the flowcharts shown in FIGS. 2, 3, 5, and 9, such processes, methods, and algorithms may be in any suitable order. It can be configured to work. In other words, the steps of the processes, methods, and algorithms described in various embodiments of the present disclosure need not be performed in the order described in this disclosure. In addition, although some steps are described as being performed asynchronously, in some embodiments these some steps may be performed simultaneously. Moreover, illustration of the process by depiction in the figures does not mean that the illustrated process excludes other changes and modifications thereto, and any of the illustrated process or steps thereof is one of the various embodiments of the present disclosure. It is not meant to be essential to more than one, nor does it mean that the illustrated process is preferred.

In operation 210, the processor 130 of the printed circuit board inspection apparatus 100 may acquire an image of the first printed circuit board on which the plurality of solder pastes are printed through the image sensor 110. The processor 130 may further obtain measurement information about the plurality of solder pastes using the obtained image. The measurement information for the plurality of solder pastes may include at least one of volume, area, height, shape, or position for each of the plurality of solder pastes.

In operation 220, when an abnormality of at least one solder paste of the plurality of solder pastes is detected using the obtained image of the first printed circuit board, the processor 130 may indicate whether the plurality of solder pastes are abnormal. You can create one image. The processor 130 determines whether the plurality of solder pastes are abnormal in order to determine a defect type of the screen printer 101 associated with at least one solder paste in which an abnormality is detected using a machine-learning based model stored in the memory 120. At least one image may be generated.

In operation 230, the processor 130 uses the machine-learning based model stored in the memory 120 to display at least one defect type representing the at least one defect type of the screen printer 101 and at least one image generated. The value can be obtained. For example, the machine-learning based model extracts features of the generated at least one image, and based on the extracted features, at least one defect type of the screen printer 101 and at least one image generated. At least one value representing relevance may be derived.

In operation 240, the processor 130 may determine a defect type associated with at least one solder paste in which an abnormality is detected among at least one defect type of the screen printer 101 based on the obtained at least one value.

For example, the processor 130 may determine a defect type corresponding to a value that is equal to or greater than a preset threshold among the acquired values as a defect type associated with at least one solder paste in which an abnormality is detected.

As another example, the processor 130 may determine a plurality of defect types corresponding to the top n values among the obtained values as defect types associated with at least one solder paste in which an abnormality is detected. Here, the number corresponding to n can be adjusted according to the setting.

3 is a flowchart of a method of detecting an abnormality of at least one solder paste printed on a printed circuit board according to various embodiments of the present disclosure.

In operation 310, the processor 130 of the printed circuit board inspection apparatus 100 uses a plurality of solders printed on the first printed circuit board by using an image of the first printed circuit board acquired through the image sensor 110. First measurement information about the pastes may be obtained. For example, the processor 130 may measure information on at least one of a volume, an area, a height, a shape, or a position of each of the plurality of solder pastes through an image, and generate first measurement information through the measured information. Can be.

In operation 320, the processor 130 may retrieve the second measurement information from the memory 120. As another example, the processor 130 may receive the second measurement information from an external server. The second measurement information may be measurement information about a plurality of solder pastes printed on each of the plurality of second printed circuit boards that are inspected by the printed circuit board inspection apparatus 100 prior to the first printed circuit board. For example, the processor 130 may check a plurality of solders printed on each of the plurality of second printed circuit boards through an image of each of the plurality of second printed circuit boards during the inspection of each of the plurality of second printed circuit boards. Information about at least one of a volume, an area, a height, a shape, or a position of the pastes may be measured, and second measurement information may be generated through the measured information. The processor 130 may store the generated second measurement information in the memory 120 or in an external server. In addition, after the inspection of the first printed circuit board is performed, the processor 130 may update the second measurement information by adding the first measurement information to the second measurement information.

In operation 330, the processor 130 may detect an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board using the first measurement information and the second measurement information. . For example, the processor 130 may determine the difference between the first measurement information and the second measurement information by comparing the first measurement information and the second measurement information, and if there is a part that is outside the set range from the determined difference, One of the plurality of solder pastes printed on the printed circuit board may detect that there is an abnormality in at least one solder paste corresponding to a portion outside the preset range. In addition, the processor 130 determines a difference between the first measurement information and the second measurement information, and when the determined difference does not deviate from a preset range, the processor 130 may have abnormalities in the plurality of solder pastes printed on the first printed circuit board. It can be detected as missing.

4A-4C illustrate a printed circuit board on which a graph showing first measurement information and second measurement information and a solder paste in which an abnormality is detected are printed according to various embodiments of the present disclosure.

4A to 4C, the second measurement information 412, 422, and 432 is measurement information for a plurality of second printed circuit boards in time series according to the order of inspection by the printed circuit board inspection apparatus 100. Can be generated in combination. Here, the measurement information may include information measured for each of the plurality of solder pastes printed on each printed circuit board.

For example, the first measurement information 411, 421, 431 and the second measurement information 412, 422, 432 are area ratios between pads on which a plurality of solder pastes printed on one printed circuit board are printed. The measured information may be included for each of the plurality of solder pastes aligned according to the " However, this is only for the purpose of description and the present invention is not limited thereto. For each of the plurality of solder pastes included in the first measurement information 411, 421, and 431 and the second measurement information 410, 420, and 430. The plurality of pieces of solder paste may be sorted according to ID information and aspect ratio of printed pads, and may be arranged in time series according to the inspection order of each of the plurality of solder pastes. It may be ordered randomly.

In one embodiment, the processor 130 of the printed circuit board inspection apparatus 100 may compare the first measurement information 411, 421, 431 and the second measurement information 412, 422, 432. The processor 130 compares the first measurement information 411, 421, 431 and the second measurement information 412, 422, 432 with the first measurement information 411, 421, 431 and the second measurement information 412. , 422, 432 can be determined. In addition, the processor 130 may determine portions 413, 423, and 433 that deviate from the set range from the determined difference. The processor 130 may detect that there is an abnormality in at least one solder paste corresponding to the portions 413, 423, and 433 outside the determined set range.

In one embodiment, the processor 130 adds the first measurement information 411, 421, 431 to the second measurement information 412, 422, 432 after the inspection of the first printed circuit board ends. The second measurement information 412, 422, 432 may be updated. For example, the processor 130 may update the second measurement information 412, 422, and 432 only when no abnormality occurs in the plurality of solder pastes printed on the first printed circuit board. As another example, the processor 130 may update the second measurement information 412, 422, 432 even when an abnormality occurs in at least one solder paste of the plurality of solder pastes printed on the first printed circuit board. have.

5 is a flowchart of a method of determining a defect type of a screen printer associated with at least one solder paste in which an anomaly is detected, according to various embodiments of the present disclosure.

In operation 510, the processor 130 of the printed circuit board inspection apparatus 100 may determine a plurality of second printed circuit boards according to an inspection order of each of the plurality of second printed circuit boards, based on the second measurement information. It is possible to determine a change in the measurement information for the plurality of printed solder pastes. For example, the processor 130 may determine how the measurement information for the plurality of solder pastes printed on each of the plurality of second printed circuit boards is changed according to the inspection order of each of the plurality of second printed circuit boards. You can judge.

In operation 520, the processor 130 may determine a defect related to at least one solder paste in which at least one defect of the screen printer 101 is detected, based on a change in measurement information regarding the plurality of determined solder pastes. The type can be determined. For example, if an abnormality of at least one solder paste printed on the first printed circuit board is detected, the processor 130 may measure measurement information on the plurality of determined solder pastes before using the machine-learning based model. Based on the change in, it is possible to determine the defect type associated with at least one solder paste in which an anomaly is detected. As another example, the processor 130 may simultaneously perform defect type determination through a machine-learning based model and defect type determination based on a change in measurement information. In this case, the processor 130 finally determines the defect type associated with at least one solder paste in which an abnormality is detected, using the defect type determination result based on the change in the measurement information and the defect type determination through the machine-learning based model. Can be.

In operation 530, the processor 130 may output information about the determined defect type through the display 140 or another output device. In addition, the processor 130 may transmit information about the determined defect type to another electronic device.

6 is a graph illustrating a plurality of pieces of information decomposed from the second measurement information and the second measurement information according to various embodiments of the present disclosure.

In one embodiment, the second measurement information is measured in time series in which the measurement information for the plurality of solder pastes printed on each of the plurality of second printed circuit boards in the order in which the inspection is performed by the printed circuit board 100. Can be generated in combination.

As illustrated in FIG. 6, when the first to ninth PCBs are inspected in time series by the printed circuit board 100 in the inspection order, the second measurement information 610 may include a plurality of printed on the first PCB. Measurement information on the solder pastes and the measurement information on the plurality of solder pastes printed on the ninth PCB may be generated by being coupled in time series according to the inspection order of the first to ninth PCBs. The second measurement information 610 may be transferred to the printed circuit board inspection apparatus 100 and then used to detect an abnormality of the plurality of solder pastes printed on the printed circuit board on which the inspection is performed.

In addition, the measurement information for the plurality of solder pastes printed on the first PCB to the measurement information for each of the plurality of solder pastes printed on the ninth PCB may include the plurality of solder pastes printed on the first to ninth PCBs. Can be sorted based on information about the pads. The measurement information for the plurality of solder pastes printed on the aligned first PCB to the measurement information for each of the plurality of solder pastes printed on the ninth PCB is time-series according to the inspection order of the first to ninth PCBs. Can be combined. For example, printed on the first PCB based on characteristics of each of the plurality of pads (eg, area ratio, aspect ratio, position cluster group ID or pad ID, etc.). The measurement information for the plurality of solder pastes to the measurement information for each of the plurality of solder pastes printed on the ninth PCB may be aligned.

Since the pads on the printed circuit board on which the plurality of solder pastes are printed may have different characteristics of each pad, in order to analyze the second measurement information 610 more accurately, the second measurement information 610 may be expressed by Equation 1 below. It may be decomposed into period information 620, trend information 630, conditional variance information 640, moving trend information 650 of an average, and error information 660.

Here, Y _t is the second measurement information 610, S _t is the period information 620, T _t is the trend information 630, h _t is the conditional variance information 640, μ _t is the moving trend information 650 of the average, and E _t is the error information 660. As such, when the second measurement information 610 is generated, the second measurement information 610 is decomposed to cycle information 620, trend information 630, conditional variance information 640, and moving trend information 650 of an average. ) And error information 660 may be generated.

For example, the period information 620 is the same on the first to ninth PCBs, and the trend information 630 indicates the tendency of the measurement information for the plurality of solder pastes printed on the first to ninth PCBs. Can be. The conditional variance information 640 is information indicating variability according to different characteristics of each pad, and the moving trend information 650 of averages is information indicating a trend of measurement information for solder paste having different average values due to different characteristics of each pad. Can be. The error information 660 may be noise obtained by removing the period information 620, the trend information 630, the conditional variance information 640, and the average moving trend information 650 from the second measurement information 610. . As such, by decomposing the second measurement information 610 in consideration of different characteristics of each pad of the printed circuit board, the measurement information of the plurality of solder pastes printed on the first PCB to the plurality of printed on the ninth PCB The measurement information on the solder pastes may more accurately analyze the second measurement information 610 which is coupled in time series according to the inspection order of the first to ninth PCBs. In addition, the defect of the screen printer 101 may be determined by determining a change in the measurement information of the plurality of solder pastes printed on the first PCB to the plurality of solder pastes printed on the ninth PCB according to the analysis result. Can be determined.

In one embodiment, the first measurement information for the plurality of solder pastes printed on the first printed circuit board also includes period information, trend information, conditional variance information, and average according to characteristics of each of the plurality of pads of the first printed circuit board. It can be decomposed into the moving trend information and error information of. The processor 130 of the printed circuit board inspection apparatus 100 may decompose period information, trend information, conditional dispersion information, average moving trend information, and error information decomposed from the first measurement information. Period information 620, trend information 630, conditional variance information 640, average moving trend information 650, and error information 660 may be compared. The processor 130 may detect whether at least one solder paste of the plurality of solder pastes printed on the first printed circuit board is abnormal according to the comparison result.

7A and 7B illustrate images indicating whether a plurality of solder pastes printed on a printed circuit board are abnormal according to various embodiments of the present disclosure.

In one embodiment, the processor 130 of the printed circuit board inspection apparatus 100 uses at least one image 720, 730 to train the machine-learning based model via the image 710 on the first printed circuit board. , 740 may be generated. The processor 130 may use the learned machine-learning based model to derive a value indicating a relationship between the generated at least one image and at least one defect type of the screen printer 101.

For example, when the machine-learning based model is trained through an image 720 indicating whether a plurality of solder pastes are printed on the first printed circuit board, the processor 130 may print the first print as shown in FIG. 7A. An image 720 indicating whether a plurality of solder pastes printed on the circuit board is abnormal may be generated. In the generated image 720, information indicating whether a plurality of solder pastes printed on the first printed circuit board is abnormal may be displayed. For example, the at least one solder paste in which no abnormality is detected among the plurality of solder pastes may be displayed in black, and the at least one solder paste in which abnormality is detected may be displayed in white. In addition, whether the plurality of solder pastes are abnormal may be indicated by a value of zero or one. For example, a pixel value corresponding to at least one solder paste for which no abnormality is detected may be set to 0, and a pixel value corresponding to at least one solder paste for which abnormality is detected may be set to 1.

As another example, a machine-learning based model includes an image 730 indicating whether there is an overpayment for a plurality of solder pastes printed on the first printed circuit board, and a plurality of solder pastes printed on the first printed circuit board. In the case of learning through the image 740 indicating whether the solder is abnormal or not, the processor 130 may check the image 730 and the plurality of solder pastes indicating the over-payment of the plurality of solder pastes, as shown in FIG. 7B. An image 740 indicating whether an abnormality is present may be generated. The generated image 730 may indicate whether the solder pastes are overcharged. The generated image 740 may indicate whether the solder is abnormal for the plurality of solder pastes.

In addition, when the machine-learning based model is trained through at least one image different from the above-described images, the processor 130 may generate at least one image different from the above-described images.

8 is a conceptual diagram for a machine learning-based model, in accordance with various embodiments of the present disclosure.

In one embodiment, the processor 130 of the printed circuit board inspection apparatus 100 includes an abnormality of a plurality of solder pastes printed on each of the plurality of printed circuit boards classified according to at least one defect type of the screen printer 101. A machine-learning based model can be trained through a plurality of images indicating whether or not. Hereinafter, for convenience of description, a description will be made based on a CNN model, which is one of the machine-learning based models, but is not limited thereto. Various machine-learning based models may be used.

In one embodiment, the image used to train the CNN model is an image indicating whether there is a problem with the plurality of solder pastes printed on the printed circuit board, and whether there is an overpayment of the plurality of solder pastes printed on the printed circuit board. At least one of an image indicating and an image indicating whether or not a solder is abnormal for a plurality of solder pastes printed on a printed circuit board may be used.

For example, the feature extraction layer 810 of the CNN, which includes at least one convolution layer and at least one pooling layer, is a feature of an image that indicates abnormality for a plurality of solder pastes. Can extract them. The feature extraction layer 810 may use a 3x3 or 5x5 filter to extract features of the image. The coefficients of the 3x3 or 5x5 filter may be determined in the course of the CNN learning through a plurality of images indicating whether or not the plurality of solder pastes printed on each of the plurality of printed circuit boards classified according to the at least one defect type.

In addition, the classifier 820 configured as a fully connected layer of the CNN may classify an image indicating whether a plurality of solder pastes are abnormal using the features extracted from the feature extraction layer 810. The classification result derived through the classifier 820 may include a value indicating an image and a relationship between at least one defect type and an image indicating whether the plurality of solder pastes are abnormal.

Although the method has been described through specific embodiments, the method may also be embodied as computer readable code on a computer readable recording medium. Computer-readable recording media include all kinds of recording devices that store data that can be read by a computer system. The computer-readable recording medium may include a ROM, a RAM, a CD-ROM, an optical data storage device, and the like. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. In addition, functional programs, codes, and code segments for implementing the above embodiments can be easily inferred by programmers in the art to which the present disclosure belongs.

While the technical spirit of the present disclosure has been described with reference to some embodiments and the examples shown in the accompanying drawings, the technical spirit and scope of the present disclosure may be understood by those skilled in the art. It will be appreciated that various substitutions, modifications, and alterations can be made in the scope. Also, such substitutions, modifications and variations are intended to be included within the scope of the appended claims.

Claims (22)

In the apparatus for inspecting a printed circuit board,
A memory for storing a machine-learning based model trained to derive at least one value representing a relationship between at least one defect type of a screen printer and an image indicating whether a plurality of solder pastes printed on a printed circuit board are abnormal; And
A processor electrically connected to the memory
Including, the processor,
When an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board is detected by using an image of the first printed circuit board, at least one of the plurality of solder pastes indicates whether the abnormality is present. Generate an image, wherein the at least one image is an image in which a first solder paste in which an abnormality is detected and a second solder paste in which no abnormality is detected among the plurality of solder pastes printed on the first printed circuit board are distinguished; ,
Input the generated at least one image into the machine-learning based model,
Obtain at least one value from said machine-learning based model indicating at least one defect type of said screen printer and an association of said generated at least one image,
And determine a defect type of the screen printer associated with at least one solder paste in which the abnormality of the at least one defect type of the screen printer is detected based on the obtained at least one value. The method of claim 1,
The processor,
Receive an image for the first printed circuit board,
Detecting an abnormality of the at least one solder paste of the plurality of solder pastes by using the received image of the first printed circuit board,
And if the abnormality of the at least one solder paste is detected, generating the at least one image. The method of claim 1,
The processor,
Detecting whether the at least one solder paste of the plurality of solder pastes is abnormal by using first measurement information of the plurality of solder pastes obtained through an image of the first printed circuit board Circuit board inspection device. The method of claim 3,
The first measurement information,
A printed circuit board inspection device comprising at least one of volume, area, height, shape or location for the solder paste. The method of claim 4, wherein
The memory,
Further storing second measurement information on the plurality of solder pastes printed on each of the plurality of second printed circuit boards, which are inspected for abnormality by the printed circuit board inspection apparatus, before the first printed circuit board,
The processor,
And detecting the abnormality of the at least one solder paste of the plurality of solder pastes using the first measurement information and the second measurement information. The method of claim 5,
The second measurement information,
A printed circuit board, in which measurement information for a plurality of solder pastes printed on each of the plurality of second printed circuit boards is combined in time series, according to an inspection order of each of the plurality of second printed circuit boards; Inspection device. The method of claim 6,
The processor,
Based on the second measurement information, the change of the measurement information for the plurality of solder pastes printed on each of the plurality of second printed circuit boards according to the inspection order of each of the plurality of second printed circuit boards is determined. and,
Determining a defect type of the screen printer associated with an abnormality of at least one solder paste printed on the first printed circuit board, among the at least one defect type of the screen printer, based on the determined change of the measurement information. Circuit board inspection device. The method of claim 1,
At least one defect type of the screen printer is
Squeeze blade defects, squeeze blade fixing defects, support defects, table defects, defects due to poor grid lock setting, defects due to neglect of solder paste and poor solder paste kneading, defects due to insufficient solder paste and poor stencil contact A printed circuit board inspection device comprising at least one of the defects. The method of claim 1,
The machine-learning based model,
A printed circuit board inspection device, a model of a convolution neural network (CNN). The method of claim 1,
The processor,
A printed circuit board learning the machine-learning based model through a plurality of images indicating whether a plurality of solder pastes printed on each of the plurality of printed circuit boards classified according to at least one defect type of the screen printer Inspection device. The method of claim 10,
The processor,
And determining the defect type of the screen printer associated with the at least one solder paste in which the abnormality is detected, and then training the machine-learning based model through the generated at least one image. The method of claim 1,
The processor,
A printed circuit board that determines a defect type of the screen printer corresponding to a value that is equal to or greater than a preset threshold among the obtained at least one value as a defect type of the screen printer associated with at least one solder paste in which the abnormality is detected. Inspection device. A non-transitory computer readable recording medium having recorded thereon a program for running on a computer, comprising:
When the program is executed by a processor, the processor,
When an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board is detected by using an image of the first printed circuit board, at least one of the plurality of solder pastes indicates whether the abnormality is present. Generating an image, wherein the at least one image is an image in which a first solder paste in which an abnormality is detected and a second solder paste in which no abnormality is detected among the plurality of solder pastes printed on the first printed circuit board are distinguished Im-;
The at least one generated in the machine-learning based model trained to derive at least one value indicative of at least one defect type of the screen printer and an image indicating whether the plurality of solder pastes printed on the printed circuit board are abnormal. Inputting an image of the;
Obtaining, from the machine-learning based model, at least one value indicative of an association of at least one defect type of the screen printer with the generated at least one image; And
Based on the obtained at least one value, determining a defect type of the screen printer associated with at least one solder paste in which the abnormality is detected among the at least one defect type of the screen printer
And a computer readable recording medium comprising executable instructions for causing the computer to perform the operation. The method of claim 13,
The executable command is:
Receiving an image for the first printed circuit board;
Detecting an abnormality of the at least one solder paste of the plurality of solder pastes by using the received image of the first printed circuit board; And
If an abnormality of the at least one solder paste is detected, generating the at least one image
Computer readable recording medium. The method of claim 13,
The executable command is:
Detecting an abnormality of the at least one solder paste of the plurality of solder pastes by using first measurement information of the plurality of solder pastes obtained through the image of the first printed circuit board.
Computer readable recording medium. The method of claim 15,
Measurement information for the at least one solder paste,
And at least one of volume, area, height, shape, or position for the at least one solder paste. The method of claim 15,
Detecting whether the at least one solder paste is abnormal among the plurality of solder pastes,
Detecting whether the at least one solder paste is abnormal among the plurality of solder pastes using the first measurement information and the second measurement information.
Including,
The second measurement information,
Measurement information of a plurality of solder pastes printed on each of the plurality of second printed circuit boards in accordance with an inspection order of each of the plurality of second printed circuit boards which have been checked for abnormality before the first printed circuit board. Computer-readable recording medium is generated by combining the time series. The method of claim 13,
The machine-learning based model,
A computer readable recording medium that is a convolution neural network (CNN) model. The method of claim 13,
The executable command is:
Training the machine-learning based model through a plurality of images indicating whether a plurality of solder pastes classified according to at least one defect type of the screen printer are abnormal;
A computer readable recording medium for further performing. The method of claim 13,
Determining the defect type,
Determining a defect type of the screen printer corresponding to at least one value of the obtained at least one value that is greater than or equal to a preset threshold value as a defect type of the screen printer associated with the sensed at least one solder paste.
And a computer readable recording medium. A method of determining a defect type of a screen printer in a printed circuit board inspection device,
When an abnormality of at least one solder paste of the plurality of solder pastes printed on the first printed circuit board is detected by using an image of the first printed circuit board, at least one of the plurality of solder pastes indicates whether the abnormality is present. Generating an image, wherein the at least one image is an image in which a first solder paste in which an abnormality is detected and a second solder paste in which no abnormality is detected among the plurality of solder pastes printed on the first printed circuit board are distinguished Im-;
The at least one generated in the machine-learning based model trained to derive at least one value indicative of an association between at least one defect type of a screen printer and an image indicating a plurality of solder pastes printed on a printed circuit board. Inputting an image of the;
Obtaining, from the machine-learning based model, at least one value indicating a relationship between at least one defect type of the screen printer and the generated at least one image; And
Based on the obtained at least one value, determining a defect type of the screen printer associated with at least one solder paste in which the abnormality is detected among the at least one defect type of the screen printer
Defect type determination method comprising a. In the apparatus for inspecting a printed circuit board,
A memory for storing a machine-learning based model trained to derive at least one value representing a relationship between at least one defect type of a screen printer and an image indicating whether a plurality of solder pastes printed on a printed circuit board are abnormal; And
A processor electrically connected to the memory
Including, the processor,
Receive an image of a first printed circuit board having a plurality of solder pastes printed thereon,
Detecting at least one first solder paste having an abnormality among the plurality of solder pastes by using an image of the first printed circuit board,
When the at least one first solder paste is detected, only at least one first solder paste having the abnormality is displayed, and at least one second solder paste having no abnormality is generated at least one image is not displayed. and,
Input the generated at least one image into the machine-learning based model,
Obtain at least one value from the machine-learning based model indicating at least one defect type of the screen printer and an association of the generated at least one image,
A based on the obtained at least one value, determining a defect type of the screen printer associated with at least one first solder paste in which the abnormality is detected among the at least one defect type of the screen printer. .

Images