JPH06194133A - Soldering appearance inspecting method - Google Patents

Abstract

PURPOSE:To judge the quality of a soldered fillet part by photographing a gloss pattern from a soldered fillet part illuminated by ring light, and performing pattern matching on this pattern on the basis of the value on a predetermined feature extraction line in relation to the previously obtained MF function of a standard pattern. CONSTITUTION:A gloss pattern from a soldered fillet part illuminated by a ring light 2 is photoraphed by a camera 1 and stored in an image memory 4. A pattern collating process 6 determines a matching MF function on this gloss pattern photographed by the camera 1, on the basis of the value within the area of a transmission type membership function and the value within the area of an inhibiting type membership function out of MF functions from a predetermined feature extraction line. The soldered state of the determined MF function is judged to be the state of the photographed gloss pattern, that is, a fuzzy neuron LS1 chip 5 inputs the gloss pattern from an image memory 4 and writes the previously obtained MF function, and outputs this MF function as the MF function matching the gloss pattern.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a soldering appearance inspection method for inspecting the appearance of soldering.

Recently, many methods of inspecting a soldering portion of an electronic component mounted on a printed circuit board with an automatic soldering appearance inspection apparatus have been announced and commercialized. Among them, many optical devices have been announced from the viewpoints of maintainability, stability, and speed.

At this time, in the optical soldering appearance inspection, it is desired to perform inspection such as quality judgment of the soldered portion with high accuracy and speed with a simple structure.

[0004]

2. Description of the Related Art Conventionally, there are the following optical inspection devices. (1) An illumination light is applied to the solder fillet portion, and the pass / fail judgment is made based on the reflection area.

(2) There is a pseudo three-dimensional system in which several stages of ring lights are sequentially applied to the solder fillet portion and judgment is made based on the change in the position of the reflected gloss. (3) The number of ring lights used in the pseudo three-dimensional system is only one as shown in FIG. A photograph is taken and the photographed image is judged by a neural network.

[0006]

The above-mentioned conventional (1)
According to the method (1), the pass / fail determination is made based on the area reflected from the solder fillet portion, so that there is a problem that the pass / fail determination of soldering is not accurate.

According to the above-mentioned conventional method (2), it is necessary to judge by the change of the position of the gloss that is reflected by switching the lights, or a high resolution RGB color camera is required. There is a problem in that it takes time to determine whether the soldering is successful or not, and a high-resolution color camera is required, which is expensive.

According to the above-mentioned conventional method (3), as shown in FIG. 7, there is a problem that the pattern of the solder fillet portion by the ring light 22 is difficult to recognize and the processing is difficult at high speed.

The present invention solves these problems.
The value on the feature extraction line of the pattern from the solder fillet portion by the ring light is pattern-matched with the previously obtained MF function of the standard pattern, and the simple processing,
The purpose is to quickly and accurately determine the quality of the soldering fillet portion.

[0010]

[Means for Solving the Problems] Means for solving the problems will be described with reference to FIG. In FIG. 1, the camera 1
Is for photographing the gloss pattern from the solder fillet portion illuminated by the ring light 2.

In the pattern matching process 6, for the gloss pattern photographed by the camera 1, the values within the area of the transparent membership function of the MF function and the area of the forbidden membership function are determined from above the predetermined feature extraction line 7. The matching MF function is determined based on the value.

[0012]

According to the present invention, as shown in FIG. 1, the camera 1 takes a picture of the gloss pattern from the solder fillet portion illuminated by the ring light 2, and the pattern matching process 7 determines the gloss pattern taken by the camera 1 in advance. A matching MF function is determined based on the value in the region of the transparent membership function of the MF function and the value in the region of the forbidden membership function from the feature extraction line 7, and the solder of the determined MF function is determined. The attached state is determined to be the state of the photographed gloss pattern.

Further, the camera 1 photographs the gloss pattern from the solder fillet portion illuminated by the ring light 2, and the pattern matching process 6 takes M from the predetermined feature extraction line 7 for the gloss pattern photographed by the camera 1.
When the F function is a monomodal transparent membership function, the inversion of the maximum value in the domain of the transparent membership function and the maximum value in the domain of the forbidden membership function is found, while the monomodal transparent member is calculated. The maximum value in the area of the transparent membership function is calculated when it is not the ship function, and the minimum value of the calculated maximum values is repeatedly calculated for all MF functions. The soldering state of the value MF function is determined as the state of the gloss pattern.

Therefore, the value on the feature extraction line 7 of the gloss pattern from the solder fillet portion by the ring light 2 is subjected to pattern matching with the MF function of the standard pattern obtained in advance to determine the soldering state,
By matching a small amount of data, it is possible to quickly and accurately determine the quality of the soldering fillet portion.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the construction and operation of an embodiment of the present invention will be described in detail with reference to FIGS.

FIG. 1 shows a block diagram of an embodiment of the present invention.
In FIG. 1, a camera 1 photographs a gloss pattern from a soldering fillet portion of a component 3 illuminated by a ring light 2. The gloss pattern obtained by this photographing is stored in the image memory 4.

The ring light 2 is for producing a gloss pattern which is reflected by a component 3 which is arranged and soldered on a substrate 31 to be inspected, for example, a soldering fillet portion of an LSI pin. By illuminating using the ring light 2, the gloss pattern changes as shown in FIGS. 2 and 3 described later in accordance with the state of the solder fillet portion (normal, unsoldered, unwet, etc.). Therefore, the present invention determines the shape of the gloss pattern from the solder fillet portion at an extremely high speed (see FIG. 6).

The component 3 is a component to be placed on the substrate 31 to be inspected and soldered, and is an LSI having pins,
It is a component such as an IC. The image memory 4 temporarily stores a gloss pattern from the solder fillet portion of the component 3 captured by the camera 1.

The fuzzy neuron LSI chip 5 inputs a gloss pattern from the image memory 4 and obtains M in advance.
The F function is written and the degree of matching between the gloss pattern and the MF function is output as a value. Here, a fuzzy neuron LSI chip 5 is provided for each MF function.
By inputting the same gloss pattern respectively and outputting the degree of matching with the MF function in parallel as a value, and determining the soldering state of the most matching MF function as the soldering state of the gloss pattern, It becomes possible to judge the soldering state of the pins of the component 3 in parallel at a very high speed. In this example, a fuzzy neuron LSI chip 5 is provided for each MF function, and the value is output to the host.
The host determines the soldering state of the pins of the component 3 based on the value, but one fuzzy neuron LSI
The values may be sequentially obtained by the chip 5, and the soldering state of the pins of the component 3 may be determined from these values.

The pattern matching process 6 is a process performed within the fuzzy neuron LSI chip 5 and is an LSI chip for obtaining a numerical value indicating the degree of matching between the gloss pattern of the pin of the component 3 and the MF function (see FIG. 6). ).

Next, a soldering state and a gloss pattern example will be described with reference to FIG. Here, the upper stage of (a) to (f) in the figure shows a side view, and the lower stage shows a plan view. FIG. 2A shows a normal state. The black part in (a) of the upper side view shows the soldered part. The half ring in (b) of the lower plan view shows the gloss pattern photographed by the camera 1 by irradiating the soldering portion of the pin of the component 3 with the ring light 2. When the gloss pattern of this half ring is obtained, the soldering state of the pins of component 3 is normal (good product).
Is.

FIG. 2B shows a state without solder. Since there is no solder in (a) of the upper side view, there is no soldered part. Since there is no solder in the lower plan view (b), there is no gloss pattern, and the soldered state of the pins of the component 3 is no solder.

FIG. 2C shows a state in which solder has not been attached. The black part in (a) of the upper side view shows the solder part. The half ring in (b) of the lower plan view shows the gloss pattern photographed by the camera 1 by irradiating the soldering portion of the pin of the component 3 with the ring light 2. When the gloss pattern of this half ring is obtained, the pins 3 of the component 3 are not soldered yet.

FIG. 2D shows a non-wetting state. The black part in (a) of the upper side view shows the solder part. The ring (b) in the lower plan view shows the gloss pattern photographed by the camera 1 by irradiating the soldering portions of the pins of the component 3 with the ring light 2. When the gloss patterns of these rings are obtained, the soldering state of the pins of the component 3 is not wet.

FIG. 2E shows an excessive solder state. The black part in (a) of the upper side view shows the soldered part. The ring (b) in the lower plan view shows the gloss pattern photographed by the camera 1 by irradiating the soldering portions of the pins of the component 3 with the ring light 2. When the gloss pattern of this ring is obtained, the soldering state of the pins of the component 3 is excessive.

FIG. 2 (f) shows an insufficient solder state. The black part in (a) of the upper side view shows the soldered part. A part of the ring (b) in the lower plan view illuminates the soldering part of the pin of the component 3 with the ring light 2 and the camera 1
The gloss pattern photographed in 1. is shown. When the gloss pattern of a part of this ring is obtained, the soldering state of the pins of the component 3 is insufficient.

Next, referring to FIG. 3, the feature line extraction line and the MF
The function example will be described in detail. FIG. 3A shows an example of normal soldering. FIG. 3A-1 shows an image of the solder fillet portion taken by the camera 1.

FIG. 3A-2 shows a partial enlargement of the gloss pattern of the solder fillet portion of FIG. 3A-1. here,
The thick line is the gloss pattern. This glossy pattern is characterized by the soldering condition of the solder fillet,
In the present invention, feature extraction lines are set on the gloss pattern without performing pattern matching on the entire surface.
In order to perform matching with the membership function only for this feature extraction line, the membership function shown on the right side is obtained in advance and registered as the membership function of the standard pattern at the time of normal soldering.

FIG. 3 (a-3) shows the membership function of the standard pattern at the time of normal soldering of FIG. 3 (a).
This is a membership function obtained by collecting a plurality of gloss patterns at the time of normal soldering and obtaining the feature extraction line set in (a-2) of FIG. 3, as shown in FIG. 5 described later ( (See the description of FIG. 5). Here, transparency represents a transparent membership function, and prohibition represents a prohibition membership function. The transparent membership function indicates the range within which the gloss pattern should be transparent. The forbidden membership function indicates the range in which the gloss pattern should not penetrate in this area. In the case of normal soldering, the feature extraction line consists of transmission and inhibition, and has a monomodal transmission type membership function. There are two feature extraction lines, transmission-1 and transmission-2, which are a bimodal transmission type membership function.

As described above, with respect to the gloss pattern from the solder fillet portion at the time of normal soldering, the feature extraction line,
, And the membership functions are obtained and registered as shown in the figure.

FIG. 3B shows an example in which no solder is attached. Figure 3
(B-1) indicates an image of the solder fillet portion taken by the camera 1.

FIG. 3B-2 shows a partial enlargement of the gloss pattern of the solder fillet portion in FIG. 3B-1. here,
The thick line is the gloss pattern. This glossy pattern is characterized by the soldering condition of the solder fillet,
In the present invention, feature extraction lines are set on the gloss pattern without performing pattern matching on the entire surface.
In order to perform matching with the membership function only for this feature extraction line, the membership function shown on the right side is obtained in advance and registered as the membership function of the standard pattern before soldering.

FIG. 3B-3 shows the membership function of the standard pattern before soldering in FIG. 3B.
This is a membership function obtained by collecting a plurality of gloss patterns when soldering is not yet performed and obtaining the feature extraction lines set in (b-2) of FIG. 3, respectively, as shown in FIG. 5 described later. is there. When soldering is not yet performed, the feature extraction line consists of prohibition and transmission, which is a monomodal transmission type membership function. There are two feature extraction lines, transmission-1 and transmission-2, which are a bimodal transmission type membership function.

As described above, with respect to the gloss pattern from the solder fillet portion when soldering is not yet performed, the feature extraction line,
, And the membership functions are obtained and registered as shown in the figure.

FIG. 3C shows an example of solder non-wetting. FIG. 3C-1 shows an image of the solder fillet portion taken by the camera 1.

FIG. 3C-2 shows a partial enlargement of the gloss pattern of the solder fillet portion of FIG. 3C-1. here,
The thick line is the gloss pattern. This glossy pattern is characterized by the soldering condition of the solder fillet,
In the present invention, feature extraction lines are set on the gloss pattern without performing pattern matching on the entire surface.
In order to perform matching with the membership function for these feature extraction lines, the membership function shown on the right side is obtained in advance and registered as the membership function of the standard pattern at the time of non-wetting with solder.

FIG. 3C-3 shows the membership function of the gloss pattern at the time of non-wetting with soldering in FIG. 3C. As shown in FIG. 5, which will be described later, a plurality of gloss patterns at the time of non-wetting with soldering are collected, and (c-2) in FIG.
The membership function obtained for each of the feature extraction lines set in. In the case of non-wetting with solder, the feature extraction line consists of transmission-1, transmission-2 and prohibition,
It is a bimodal transparent membership function. There are two feature extraction lines, transmission-1 and transmission-2, which are bimodal transmission-type membership functions. The feature extraction line is a Don't care which is not a target for matching here.

As described above, with respect to the gloss pattern from the solder fillet portion when the soldering is not wet, the feature extraction line,
For each, the membership function is obtained and registered as shown.

FIG. 4 shows an example of the gloss pattern in the normal state of the present invention. This shows an example of the gloss pattern at the time of normal soldering of FIG. (A) shows a gloss pattern when the entire lead portion is soldered uniformly. (B) shows a gloss pattern when the amount of solder is a little smaller than (a). (C) shows a gloss pattern when the amount of solder is smaller than that in (b) and the amount of solder is slightly asymmetric. All of these gloss patterns are normal gloss patterns, and these normal gloss patterns are determined according to the flowchart of FIG. 6 described later.

Next, referring to FIG. 5, (a-3) in FIG.
The procedure for calculating and registering the membership functions (MF functions) of (b-3) and (c-3) will be described in detail.
In FIG. 5, S1 collects a plurality of patterns in the same category. This collects a plurality of patterns of the same category of the solder fillet portion, for example, a plurality of normal gloss patterns of FIG. 4 described above.

In step S2, the feature extraction line is set. this is,
As described above with reference to FIG. 3A, a feature extraction line for matching gloss patterns, here, a feature extraction line, is set to each of the plurality of gloss patterns collected in S1.

In step S3, the frequency of the gloss pattern is calculated.
This is the feature extraction line of the gloss pattern set in S2,
, For example, the frequency on the feature extraction line of (a) on the right side is obtained as shown in (b) on the right side.

In step S4, a membership function (MF function) of each characteristic line is created and saved. This is S
Regarding the frequency of (b) on the right side, which is the plurality of gloss patterns on the feature extraction line calculated in step 3, connect a part that is larger than a certain upper side threshold to the upper side, and connect the zero frequency part from the shoulder of this upper side. Create a shape with the width of the hypotenuse, create this as a membership function (MF function), and save it.

As described above, a plurality of gloss patterns of the same category (for example, normal, no solder, unsoldered, non-wetting, excessive solder, insufficient solder, etc.) are collected, and characteristic extraction lines of these gloss patterns, Then, the membership function is calculated from this frequency. Then, the membership function (MF function) is registered in association with the category.

Next, the procedure for finding the membership function that matches the gloss pattern of the solder fillet portion and automatically determining the soldering state will be described in detail according to the order shown in the flowchart of FIG.

In FIG. 6, in step S11, the gloss pattern of the solder fillet portion is fetched. This captures a gloss pattern obtained by photographing the soldering state of the pins of the component 3 on the board 31 to be inspected in FIG. 1 with the camera 1.

In step S12, it is determined whether or not there is one peak. This determines whether the membership function of the standard pattern at the top saved in FIG. In the case of YES (single-peaked), the process proceeds to S13. On the other hand, if NO (if not one-peaked), the process proceeds to S14.

In S13, it is determined that the answer is YES in S12, so that the maximum value of the one-peak transmissive membership function is calculated. The inversion of the maximum value of the forbidden membership function is calculated. This is, for example, within the region of the single-peaked transmission type membership function (= within the region described as transmission) of the membership functions of the feature extraction line of (a-3) in the normal state of FIG.
The maximum value of the gloss pattern corresponding to is calculated, and the inverted value of the maximum value of the gloss pattern corresponding to the area of the prohibited membership function (= the area described as prohibited) is calculated.

In S14, since NO in S12 reveals that it is not monomodal, the maximum value of the membership function is calculated. This is a gloss pattern corresponding to, for example, the area of the transmission type membership function (= the area described as transmission-1) of the membership function of the feature extraction line of (a-3) in the normal state of FIG. And the maximum value of the gloss pattern corresponding to the area of the transmission type membership function (= the area described as transmission-2).

In step S15, the minimum output value within one pattern is output. This outputs the minimum value of the values calculated in S13 or S14 as the minimum value representing the one pattern.

S16 is a matching process (matching process from S12 to S15) for all standard patterns (standard membership functions such as normal, no solder, soldering, non-wetting, excessive solder, insufficient solder stored in FIG. 5). ) Is the end. In the case of YES, it progresses to S17. N
In the case of O, the matching process from S12 to S15 is repeated for the next standard pattern (next standard membership function).

In S17, the pattern having the maximum value among all the pattern outputs is determined as matching. This is each standard pattern output in S15 (standard membership function)
Of the minimum value after the matching process with, the maximum value of the standard pattern (standard membership function) soldering state (normal, no solder, soldering, non-wetting, excessive soldering,
(Excessive amount of solder, etc.) is determined as the soldering state of the gloss pattern captured in S11.

As described above, regarding the gloss pattern photographed by the camera 1 at that time by illuminating the solder fillet portion of the pin of the component 3 with the ring light 2 of FIG. 1, the membership of all the soldering states stored in advance in FIG. A function and matching process (S12 to S15) are performed, and the soldering state of the maximum value of the minimum values of the standard patterns (standard membership function) is detected. With this, the feature extraction line of the gloss pattern, the maximum value in the area of each transmission type membership function above, and the maximum value in the area of the prohibition type membership function are inverted to calculate the minimum value in the pattern. After determining, the soldering state of the maximum value of all patterns is determined. At this time, the matching process is executed in parallel for each standard pattern (standard membership function) using a plurality of fuzzy neuron LSI chips 5 as shown in FIG. 1 described above to obtain the minimum value of each standard pattern. Since the maximum value is detected as the soldering state, the soldering state of the solder fillet portion can be detected extremely quickly and accurately.

[0054]

As described above, according to the present invention,
M of the pattern obtained in advance for the value on the feature extraction line 7 of the pattern from the solder fillet portion by the ring light 2
Since the configuration for determining the soldering state by performing the pattern matching with the F function is adopted, the pass / fail of the solder fillet portion can be quickly and accurately determined by the matching of a small amount of data. At this time, for the standard gloss pattern of the solder fillet (normal, no solder, unsoldered, non-wetting, excessive solder, insufficient solder, etc.), the feature extraction line and the above membership function (MF function) are obtained in advance. Registered, and the gloss pattern taken by the camera 1 of the solder fillet portion such as the pin of the component 3 to be inspected is registered by the membership function and the matching process (S12 to S15 in FIG. 6) of each standard pattern. Value is calculated in parallel, and by detecting the soldering state of the maximum value among them, simple matching processing, extremely fast, parallel, and highly accurate soldering state of the solder fillet part is automatically detected and inspected. It becomes possible to do.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is an example of a soldering state and a gloss pattern of the present invention.

FIG. 3 is an example of a feature extraction line and an MF function according to the present invention.

FIG. 4 is an example of a gloss pattern at a normal time of the present invention.

FIG. 5 is a flowchart of storing a membership function of the present invention.

FIG. 6 is a pattern matching process flowchart of the present invention.

FIG. 7 is an explanatory diagram of an inspection with a ring light.

[Explanation of symbols]

 1: Camera 2: Ring light 3: Component 31: Board to be inspected 4: Image memory 5: Fuzzy neuron LSI chip 6: Pattern matching process 7: Feature extraction line ,,: Example of feature extraction line

Claims (2)

[Claims] 1. A soldering appearance inspection method for inspecting the appearance of soldering, comprising: a camera (1) for photographing a gloss pattern from a solder fillet portion illuminated by a ring light (2); and a camera (1) for photographing the glossy pattern. Regarding the gloss pattern, pattern matching processing for matching is performed based on the value in the area of the transparent membership function of the MF function and the value in the area of the forbidden membership function from the predetermined feature extraction line (7) ( 6), and the soldering appearance inspection method is characterized in that the soldering state of the MF function determined to be matched by the pattern matching process (6) is determined to be the state of the photographed gloss pattern. . 2. A soldering appearance inspection method for inspecting the appearance of soldering, comprising: a camera (1) for photographing a gloss pattern from a solder fillet portion illuminated by a ring light (2); and a camera (1) for photographing. For the gloss pattern, the MF function is 1 from the predetermined feature extraction line (7).
For a peak-type transparent membership function, the maximum value in the transparent membership function area and the inverted maximum value in the forbidden membership function area are obtained.
Pattern matching processing (6) for obtaining the maximum value in the area of the transparent membership function when it is not the one-peaked transparent membership function, and the soldering state of the MF function having the minimum value among these calculated values Is determined to be the state of the gloss pattern.