JP2019100753A - Printed circuit board inspection device and printed circuit board inspection method - Google Patents

Abstract

To inspect a printed circuit board with higher accuracy than that by a conventional imaging method with a camera.SOLUTION: A printed circuit board inspection device 1 comprises: an optical displacement measuring module 2 for causing light with which a measurement object surface Wa of a printed circuit board W is irradiated and the measurement object surface Wa to be relatively scanned, and measuring the height displacement of the measurement object surface Wa and the displacement of reflected light on the basis of the scanned position of light with which the measurement object surface Wa is irradiated and the imaged position on an image sensor and the quantity of light; a feature quantity acquisition unit 3 for acquiring information on the height displacement and the displacement of reflected light as a feature quantity; a three-dimensional displacement distribution data generation unit 4 for generating three-dimensional displacement distribution data on the height and reflected light on th basis of the feature quantity; a specific region data extraction unit 5 for extracting specific region data that corresponds to an evaluation item from the three-dimensional displacement distribution data; and a determination unit 8 for applying a learned neural network to the extracted specific region data and determining whether the printed circuit board W is good or not.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a printed circuit board inspection apparatus and a printed circuit board inspection method for inspecting electronic components, soldering, and the like of a printed circuit board after mounting of the printed circuit board and electronic components in a production process of the printed circuit board.

  In recent years, everything is electronically controlled, such as mobile phones, home appliances, automobiles, etc., and high reliability to printed circuit boards is required. For this reason, in each manufacturing process of a printed circuit board, a more reliable inspection apparatus for quality control is required.

  Conventionally, as an inspection method of a printed circuit board in this type of inspection apparatus, an imaging method using a camera is known as disclosed in, for example, Patent Document 1 and Patent Document 2 below.

  As an imaging method by a camera, there is an appearance inspection method of a printed circuit board in which an image taken by a camera is judged by a computer. However, it is difficult to judge a soldered joint from an image. A plurality of images of the same part are taken, and determination is made from a plurality of images.

JP, 2015-026287, A JP, 2012-028460, A

  The conventional camera imaging method including the above-described Patent Document 1 and Patent Document 2 has the advantage that the measurement time is short, and when the light is irregularly reflected on the solder surface of the printed circuit board, the appearance of the image plane is obtained by the camera Can. However, when the solder surface of the printed circuit board is a mirror surface and light is specularly reflected, there is a problem that it is difficult to obtain the appearance of the image plane with a camera. Moreover, since a two-dimensional flat image is used, information in the height direction of the solder portion of the printed circuit board can not be acquired, and the quality determination of the printed circuit board can not be performed with high accuracy.

  By the way, in recent years, with the printed circuit board, the miniaturization of parts and the densification of mounting have been dramatically advanced, and the inspection technology for guaranteeing the quality of the printed circuit board can not follow this change. In addition, in the case where a failure occurs in a part, it often has to be discarded, which causes a large loss.

  Therefore, the present invention has been made in view of the above problems, and provides a printed circuit board inspection apparatus and a printed circuit board inspection method capable of inspecting a printed circuit board with higher accuracy than a conventional camera imaging method. The purpose is that.

In order to achieve the above object, the printed circuit board inspection apparatus according to claim 1 of the present invention is a printed circuit board inspection apparatus 1 for inspecting a printed circuit board W,
A light source 21a for outputting light, a light projection lens 21b for narrowly narrowing the light radiated from the light source and irradiating the surface to be measured Wa of the printed circuit board, and condensing reflected light or scattered light from the surface to be measured A light receiving lens 21c, and an image sensor 21d to which the light spot collected by the light receiving lens is irradiated;
A scan control unit 22 for relatively scanning the light irradiated to the surface to be measured and the surface to be measured;
A light displacement for measuring at least one of a height displacement amount of the surface to be measured or a displacement amount of a reflected light amount based on a scanning position of light irradiated to the surface to be measured and an imaging position and light amount on the image sensor A light displacement measurement module 2 including a measurement unit 23;
A feature amount acquisition unit 3 that acquires information of at least one of the height displacement amount and the displacement amount of the reflected light amount as a feature amount;
A three-dimensional displacement distribution data generation unit 4 that generates at least one of three-dimensional displacement distribution data of height or three-dimensional displacement distribution data of reflected light quantity based on the feature amount;
A specific area data extraction unit 5 for extracting specific area data corresponding to an evaluation item from the three-dimensional displacement distribution data;
The specific area data extracted by the specific area data extraction unit is applied to a learned neural network to determine the quality of the printed circuit board.

The printed circuit board inspection apparatus according to claim 2 is the printed circuit board inspection apparatus according to claim 1.
The evaluation item is the determination of whether the soldering is good or not, and the specific area is extracted from the three-dimensional displacement distribution data from the three-dimensional displacement distribution data using the specific area as the soldering area, and the solder defect is evaluated.

The printed circuit board inspection apparatus according to a third aspect of the present invention is the printed circuit board inspection apparatus according to the first aspect, wherein
The evaluation item is quality determination of component mounting, and the specific region is extracted from the three-dimensional displacement distribution data using the specific region as a component mounting region, and the defect of the component mounting is evaluated. Do.

The printed circuit board inspection apparatus according to claim 4 is the printed circuit board inspection apparatus according to any one of claims 1 to 3,
A learning data storage unit 6 which stores the quality of the inspection result by the inspector and the specific area data as learning data;
And a learning execution unit that executes learning of the neural network using the learning data.

The printed circuit board inspection method according to claim 5 is a printed circuit board inspection method for inspecting the printed circuit board W,
Irradiating the light emitted from the light source 21a to the measurement target surface Wa of the printed circuit board with a narrow aperture;
Collecting the reflected light or the scattered light from the surface to be measured and irradiating the collected light spot to the image sensor 21d;
Relatively scanning the light irradiated to the surface to be measured and the surface to be measured;
Measuring at least one of a height displacement amount of the surface to be measured and a displacement amount of a reflected light amount based on a scanning position of light irradiated to the surface to be measured and an imaging position and light amount on the image sensor; ,
Acquiring at least one information of the height displacement amount or the displacement amount of the reflected light amount as a feature amount;
Generating at least one of height three-dimensional displacement distribution data or three-dimensional displacement distribution data of reflected light quantity based on the feature amount;
Extracting specific area data corresponding to an evaluation item from the three-dimensional displacement distribution data;
Applying the above specific area data to a learned neural network to determine whether the printed circuit board is good or bad.

  According to the present invention, in the inspection of printed circuit boards and printed circuit boards after mounting electronic components, serious defects (for example, missing parts) and quality defects (for example, distortion or deviation of parts, unsold, small amount of solder, solder bridge, etc.) It is possible to determine the quality of the inspection with higher accuracy.

It is a block diagram of a printed circuit board inspection device concerning the present invention. It is a block diagram which shows the other structure of the printed circuit board test | inspection apparatus which concerns on this invention. It is a principle view of an optical displacement measurement module used for a printed circuit board inspection device concerning the present invention. It is a figure which shows an example of the scanning direction of the light irradiated to the measurement object surface of a printed circuit board. It is a flow chart which shows an outline of a printed circuit board inspection method concerning the present invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

  The printed circuit board inspection apparatus and the printed circuit board inspection method according to the present invention inspect the quality of the printed circuit board in the production process of the printed circuit board and the electronic components of the printed circuit board after mounting of the electronic component such as soldering.

  As shown in FIG. 1, the printed circuit board inspection apparatus 1 (1A) of this embodiment includes a light displacement measurement module 2, a feature amount acquisition unit 3, a three-dimensional displacement distribution data generation unit 4, a specific area data extraction unit 5, A learning data storage unit 6, a learning execution unit 7, a determination unit 8, and a determination result output unit 9 are provided.

  The light displacement measurement module 2 is a displacement sensor to which the triangulation method is applied, and includes an optical module 21, a scan control unit 22, and a light displacement measurement unit 23 as shown in FIG. 3.

  As shown in FIG. 3, the optical module 21 is a light source 21 a such as a semiconductor laser that outputs light, and a light projection that narrows the light from the light source 21 a and irradiates the measurement target surface Wa of the printed circuit board (measurement object) W The position and brightness of the image of the light spot are irradiated with the lens 21b, the light receiving lens 21c for collecting the reflected light and scattered light from the measurement target surface Wa of the printed substrate W, and the light spot collected by the light receiving lens 21c. And an image sensor 21d that outputs a current corresponding to the image data.

  In the light displacement measurement module 2 of FIG. 3, a light spot can be formed on the surface to be measured Wa by narrowing down and irradiating the laser emitted from the light source 21a with the light projection lens 21b on the surface to be measured Wa of the printed substrate W . The light reflected or scattered from the light spot is collected by the light receiving lens 21c to form an image of the light spot on the image sensor 21d. The image of the light spot moves on the image sensor 21d when the measurement surface Wa is displaced. A current corresponding to the position and brightness of the image is output from the image sensor 21d, and the output current is voltage-converted to generate a signal of sum and difference, and division is performed to perform the correction process to thereby obtain the height of the surface Wa to be measured. And the amount of displacement of the amount of reflected light (regular reflection or scattering) can be detected. Then, the scanning control unit 22 scans the light irradiated to the measurement object surface Wa of the printed circuit board W with respect to the measurement object surface Wa of the printed circuit board W, and the light displacement measurement unit 23 prints according to the scanning position It is possible to obtain information on the amount of displacement of the height of the measurement target surface Wa of the substrate W and the amount of displacement of the amount of reflected light.

  Further, as the light displacement measurement module 2, if the polygon scanner is rotated and the laser is scanned from the light source (semiconductor laser) to move the measurement position, and further, the image sensor is moved in the direction orthogonal to the laser scanning direction. It is possible to measure the amount of displacement of the height of the measurement target surface Wa of the printed circuit board W within the range scanned by the polygon scanner and the amount of displacement of the amount of reflected light.

  Although not shown, if the light displacement measurement module 2 is configured to receive regular reflection light and scattered light from the measurement target surface Wa of the printed circuit board W by two independent image sensors, the light displacement measurement module 2 responds to the amount of reflected light. Image sensor can be selected and used.

  The scan control unit 22 scans and controls the measurement target surface Wa by irradiating the measurement target surface Wa of the print substrate W with light irradiated to the measurement target surface Wa of the print substrate W. For example, as shown in FIG. 4, the entire measurement target surface Wa is scan-controlled from the left side of the measurement target surface Wa of the printed circuit board W in the order of 1 → 2 →. The scanning control unit 22 irradiates the measurement target surface Wa of the printed circuit board W, such as a method of scanning an optical system including the light source 21a and the light projection lens 21b and a method of scanning the measurement target surface Wa of the printed circuit board W. The light source and the surface Wa to be measured may be displaced relative to each other, and any method is applicable.

  The light displacement measuring unit 23 determines at least one of the height displacement amount of the measurement target surface Wa of the printed circuit board W and the displacement amount of the reflected light amount based on the information of the current output from the image sensor 21 d and the scanning information of the scanning control unit 22. taking measurement.

  The feature amount acquisition unit 3 acquires information of at least one of the height displacement amount and the displacement amount of the reflected light amount as a feature amount. The height displacement amount acquired as the feature amount and the displacement amount of the reflected light amount are each position of the measurement target surface Wa with respect to the light irradiated to the measurement target surface Wa of the printed circuit board W obtained by the scan control of the scan control unit 22 Each corresponds to the information.

  The three-dimensional displacement distribution data generation unit 4 generates at least one of the three-dimensional displacement distribution data of the height displacement amount and the three-dimensional displacement distribution data of the reflected light amount based on the feature amount acquired by the feature amount acquisition unit 3. The three-dimensional displacement distribution data is generated as an image representing the amount of height displacement and the amount of reflected light as a difference in color.

  The specific area data extraction unit 5 extracts specific area data corresponding to the evaluation item from the one or more three-dimensional displacement distribution data (height displacement amount, reflected light quantity) generated by the three-dimensional displacement distribution data generation unit 4 Do. The evaluation items are the determination of whether the soldering of each component of the printed circuit board W is good or not, or the determination of whether the component mounting of the printed circuit board W is good or not.

  The specific area is easily specified in advance from the mounting mount data on the printed circuit board W (data indicating the position of each component mounted on the printed circuit board W) by specifying in advance the soldering area and the component mounting area. Can.

  The specific region can also be specified from the three-dimensional displacement distribution data of the height displacement amount generated by the three-dimensional displacement distribution data generation unit 4 or the three-dimensional displacement distribution data of the reflected light quantity.

  The learning data storage unit 6 stores the pass / fail information of the inspection result of the printed circuit board W and the specific area data extracted by the specific area data extracting unit 5 as learning data. In addition, the pass / fail information of the inspection result of the printed circuit board W may be acquired in advance as inspection result information provided with a normal flag from inspection results of the printed circuit board W performed by the inspector and inspection result information provided with an abnormal flag. it can.

  The learning execution unit 7 executes learning of a neural network using the learning data stored in the learning data storage unit 6.

  Here, if a convolutional neural network is used in the learning execution unit 7, high-quality judgment can be made. Convolutional neural networks solve the problem of over-learning that generalization performance does not improve, which is a problem of conventional fully coupled neural networks.

  A back propagation method is often used as a learning method for multi-layered neural networks. The error back propagation method is a method in which the error of the network (the difference between the target output and the actual output) with respect to the input sample is propagated back from the input layer, and the gradient of the weight of each layer is calculated. In this calculation, as the depth goes away from the input layer, a phenomenon called gradient disappearance problem occurs in which the gradient is rapidly reduced or rapidly increased and diverges. This has made it difficult to learn multi-layered networks.

  A convolutional neural network has a structure in which layers having a function of performing specific image processing are stacked, so that the connection between layers is a sparse network. For this reason, it becomes possible to learn a multilayer network using a convolutional neural network using an error back propagation method. The configuration of this network layer is very similar to the mechanism of human vision, so optimization of learning, such as how to give learning data and the amount and quality of learning data, enables high-quality judgment as to whether it is human's cognitive ability It becomes.

  The determination unit 8 applies a learned neural network to the specific area data extracted by the specific area data extraction unit 5 to determine whether the printed circuit board W is good or bad. More specifically, in the case where the determination unit 8 determines the solderability of the printed circuit board W with the specific area as the solder area, the neural network has already learned in the soldering area extracted as the specific area data from each three-dimensional displacement distribution data. A network is applied, and it is determined whether or not there is a solder defect such as unsold, poor solder, excessive solder, joint failure, solder bridge, solder crack and the like.

  In addition, when the determination unit 8 determines the quality of the component mounting of the printed circuit board W with the specific region as the component mounting region, the neural network in the component mounting region extracted as the specific region data from each three-dimensional displacement distribution data is learned. Apply to determine the presence or absence of component mounting defects such as component loss or component displacement.

  The determination result output unit 9 outputs the result of the presence or absence of the solder defect of the printed circuit board W determined by the determination unit 8 and the result of the presence or absence of the defect in the implementation of the component.

  Next, a printed circuit board inspection method for inspecting the printed circuit board W using the printed circuit board inspection apparatus 1 configured as described above will be described with reference to FIG.

  When the printed circuit board W is inspected, first, light is emitted from the light source 21a of the light displacement measurement module 2 toward the measurement target surface Wa of the printed circuit board W while being scanned by the scan control unit 22 (ST1).

  To explain further, the scanning control unit 22 irradiates light emitted from the light source 21a of the light displacement measuring module 2 to the surface to be measured Wa of the printed circuit board W to scan the surface to be measured Wa of the printed circuit board W with the scanning width H The light displacement measurement module 2 measures the light displacement measurement module 23 so that the light displacement measurement unit 23 can measure the amount of displacement of the height of the measurement object surface Wa and the amount of reflected light from the measurement object surface Wa over the entire measurement object surface Wa. The scan control of the light irradiated to As a result, the light emitted from the light source 21a of the light displacement measurement module 2 is reflected or scattered by the surface to be measured Wa and collected on the image sensor 21d.

  Here, the focused image moves to a different position on the image sensor 21d when the distance between the light displacement measurement module 2 and the measurement target surface Wa is displaced. The amount of displacement of the height of the surface Wa to be measured can be measured from the change in the position of the focused image.

  Further, the image sensor 21d receives the collected image and outputs a current proportional to the brightness of the collected image. Therefore, the amount of reflected light (the intensity of the reflected or scattered light) can be measured.

  Next, the feature amount acquisition unit 3 calculates the feature amount information of the height displacement amount and the reflected light amount corresponding to each position information of the measurement target surface Wa of the printed circuit board W obtained by the scan control of the scan control unit 22. Acquire as (ST2).

  Next, the three-dimensional displacement distribution data generation unit 4 generates three-dimensional displacement of the height based on the feature amount (displacement information of the height and reflected light amount information corresponding to each position information of the surface to be measured). Displacement distribution data and three-dimensional displacement distribution data of reflected light quantity are generated (ST3).

  Then, the specific area data extraction unit 5 extracts data in one or more specific areas from each three-dimensional displacement distribution data generated by the three-dimensional displacement distribution data generation unit 4 (ST4).

  Then, the determination unit 8 uses the learning data stored in the learning data storage unit 6 for the specific region data extracted by the specific region data extraction unit 5 to the learned neural network learned by the learning execution unit 7. The application is performed to determine the presence or absence of component mounting or the quality of soldering (ST5), and the result of quality is output from the determination result output unit 9.

  When the determination unit 8 determines whether the soldering of the printed circuit board W is good or bad, a specific region extracted from the three-dimensional displacement distribution data is set as a soldering region. Thereby, the determination unit 8 applies the learned neural network to the soldering region extracted as the specific region data, and determines the solder defect such as unsold, insufficient solder, excessive solder, joint failure, solder bridge, solder crack, etc. And the pass / fail result is output from the decision result output unit 9.

  When the determination unit 8 determines the quality of component mounting of the printed circuit board W, the specific region extracted from the three-dimensional displacement distribution data is defined as a component mounting region. As a result, the determination unit 8 applies the learned neural network to the component mounting area extracted as the specific area data, determines a defect in component mounting such as component omission or component deviation, and determines a pass / fail result from the determination result output unit 9 Output.

  By the way, in the embodiment described above, it is also possible to use a learned neural network for extraction of a specific area. The configuration in this case is shown in FIG. In FIG. 2, components having the same or equivalent functions as those in FIG. 1 are denoted by the same reference numerals, and the description thereof is omitted.

  In the printed circuit board inspection apparatus 1 (1B) of FIG. 2, the printed circuit board W is generated from the three-dimensional displacement distribution data of the displacement amount of height generated by the three-dimensional displacement distribution data generation unit 4 and the three-dimensional displacement distribution data of the reflected light quantity. The shape data of the lead portion of each upper part is stored in the shape data storage unit 11. Then, the learning execution unit 12 executes learning of the specific region extraction neural network using the shape data stored in the shape data storage unit 11.

  The specific area extraction neural network learns in advance the shape features of the lead portion of the part, and the three-dimensional displacement distribution data generated by the three-dimensional displacement distribution data generation section 4 has been learned by the specific area data extraction section 5 By processing with a neural network, it is possible to extract the area of only the lead portion as a specific area.

  As described above, according to the present embodiment, in the inspection of the printed circuit board and the printed circuit board after mounting the electronic component, information on the height direction of the component and the solder portion of the printed circuit board can also be obtained. As compared with the above, it is possible to perform the quality determination of the inspection of the serious defect (for example, missing part etc.) or the quality defect (for example, distortion or deviation of the part, unsold solder, small amount of solder, solder bridge etc.) with high accuracy.

  Although the best mode of the printed circuit board inspection apparatus and the printed circuit board inspection method according to the present invention has been described above, the present invention is not limited by the description and the drawings according to this mode. That is, it is needless to say that all other forms, examples, operation techniques and the like made by those skilled in the art based on this form are included in the scope of the present invention.

1 (1A, 1B) Printed circuit board inspection device 2 light displacement measurement module 3 feature amount acquisition unit 4 three-dimensional displacement distribution data generation unit 5 specific region data extraction unit 6 learning data storage unit 7 learning execution unit 8 judgment unit 9 judgment result output Section 11 Shape data storage section 12 Learning execution section 21 Optical module 21a Light source 21b Projection lens 21c Light receiving lens 21d Image sensor 22 Scan control section 23 Optical displacement measurement section W Printed circuit board (measurement object)
Wa Target surface H scan width

Claims (5)

A printed circuit board inspection apparatus (1) for inspecting a printed circuit board (W)
A light source (21a) for outputting light, a light projection lens (21b) for narrowing the light radiated from the light source and irradiating the surface to be measured Wa of the printed circuit board, reflected light or scattered light from the surface to be measured A light receiving lens (21c) for collecting light, and an image sensor (21d) to which the light spot collected by the light receiving lens is irradiated;
A scan control unit (22) for relatively scanning the light irradiated to the surface to be measured and the surface to be measured;
A light displacement for measuring at least one of a height displacement amount of the surface to be measured or a displacement amount of a reflected light amount based on a scanning position of light irradiated to the surface to be measured and an imaging position and light amount on the image sensor A light displacement measurement module (2) including a measurement unit (23);
A feature amount acquisition unit (3) that acquires information of at least one of the height displacement amount and the displacement amount of the reflected light amount as a feature amount;
A three-dimensional displacement distribution data generation unit (4) that generates at least one of three-dimensional displacement distribution data of height or three-dimensional displacement distribution data of reflected light quantity based on the feature amount;
A specific area data extraction unit (5) for extracting specific area data corresponding to an evaluation item from the three-dimensional displacement distribution data;
The specific area data extracted by the specific area data extraction unit is applied to a learned neural network to determine the quality of the printed circuit board; apparatus. The evaluation item is a judgment on the quality of soldering, and the specific area in the soldering area is extracted from the three-dimensional displacement distribution data using the specific area as the soldering area, and the solder defect is evaluated. The printed circuit board inspection device according to Item 1. The evaluation item is quality determination of component mounting, and the specific region is extracted from the three-dimensional displacement distribution data using the specific region as a component mounting region, and the defect of the component mounting is evaluated. The printed circuit board inspection device according to claim 1. A learning data storage unit (6) for storing pass / fail of inspection results by the inspector and the specific area data as learning data;
The printed circuit board inspection apparatus according to any one of claims 1 to 3, further comprising: a learning execution unit (7) that executes learning of the neural network using the learning data. A printed circuit board inspection method for inspecting a printed circuit board (W), comprising:
Irradiating the light emitted from the light source (21a) to the surface to be measured (Wa) of the narrow stop of the printed circuit board;
Collecting the reflected light or the scattered light from the surface to be measured and irradiating the collected light spot to the image sensor (21d);
Relatively scanning the light irradiated to the surface to be measured and the surface to be measured;
Measuring at least one of a height displacement amount of the surface to be measured and a displacement amount of a reflected light amount based on a scanning position of light irradiated to the surface to be measured and an imaging position and light amount on the image sensor; ,
Acquiring at least one information of the height displacement amount or the displacement amount of the reflected light amount as a feature amount;
Generating at least one of height three-dimensional displacement distribution data or three-dimensional displacement distribution data of reflected light quantity based on the feature amount;
Extracting specific area data corresponding to an evaluation item from the three-dimensional displacement distribution data;
And D. applying the above specific area data to a learned neural network to determine the quality of the printed circuit board.

Images