JPH0291505A - Inspecting device for solder surface of circuit board - Google Patents

Abstract

PURPOSE:To increase decision making efficiency by picking up an image of a soldering part and generating image information, extracting the feature quantity of the soldering part from the image information, and finding an evaluation quantity for deciding whether or not soldering is normal from the feature quantity by fuzzy reasoning. CONSTITUTION:A TV camera 1 and an image memory 2 correspond to an image information generating means 3 and a feature quantity extracting means 4 extracts the feature quantity of the soldering part of an electronic component from the obtained image information. A fuzzy reasoning means 5 uses the extracted feature quantity as a variable to output the reasoning result of the normal/abnormal soldering state as the evaluation quantity for deciding the soldering state. A decision means 7 outputs the final decision result of the soldering state according to the evaluation quantity. A membership function setting means 6 suppresses, for example, the ration at which the result is neither normal nor abnormal, namely, belongs to what is called gray as the evaluation quantity to determine a membership function used by the fuzzy reasoning means 5 so as to increase the decision rate. Consequently, the ratio at which the result belongs to evaluation quantities which are neither normal nor abnormal is decreased.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application This invention relates to a circuit board solder surface inspection device used for determining the quality of soldering of electronic components mounted on a circuit board.

(b) Prior Art Conventionally, systems for automatically mounting electronic components onto a circuit board have been provided with an inspection device for inspecting the circuit board after electronic component mounting has been completed by some method.

As a method for inspecting a circuit board on which electronic components are mounted, in addition to applying a test signal to the test terminal of the circuit board and checking the operation of the circuit board, there is also a method of directly inspecting the mounting state of electronic components on the circuit board. is used.

Conventional general inspection equipment used for such applications obtains image information by capturing a planar image of the circuit board, and performs pattern matching with standard image information to check the mounting of electronic components. Inspecting for leaks or misalignment.

fc) Problems to be Solved by the Invention However, although conventional inspection devices are mainly effective for checking for mounting omissions and misalignment of electronic components in a short time, It was not very effective for checking the soldering condition of the parts. This is because the quality of soldering cannot be determined only by the similarity between the solder shape of the soldered portion of the part to be inspected and the solder shape of a standard soldered portion.

SUMMARY OF THE INVENTION An object of the present invention is to provide a circuit board solder surface inspection apparatus that can solve the above-mentioned problems by inferring the quality of soldering from the characteristic I of the soldered portion of an electronic component.

(d) Means for Solving the Problems The circuit board solder surface inspection apparatus of the present invention includes means for generating image information by capturing an image of a soldered part of an electronic component and an electronic component from this image information. The present invention is characterized by comprising: means for extracting feature quantities of the soldering portion; and fuzzy inference means for obtaining an evaluation quantity for determining the quality of soldering of an electronic component by fuzzy inference from the feature quantities.

(e) Function The circuit board solder surface inspection apparatus of the present invention first captures an image of the soldered portion of an electronic component to generate image information, and extracts the feature amount of the soldered portion of the electronic component from this image information. do.

This feature quantity is input to the fuzzy inference means as a variable.

As is well known, the fuzzy inference means is composed of a fuzzy operation section that performs fuzzy operations and a defuzzify section that performs definite value operations.The fuzzy operation section is equipped with a membership function generator that follows predetermined fuzzy rules, and It calculates the membership value (belonging value) for the variable to be used, and outputs the inference value calculated based on the result to the defuzzifier. The fuzzy rule is 1
f(xI=^and xz=8...-) then(y
= Z), (xl=A and xg=B
-・-) is called the antecedent part, and (y=Z) is called the consequent part. This fuzzy rule is determined empirically based on the characteristic amount of the soldered part and the actual quality of soldering.

FIG. 13 is a diagram for explaining one known method of outputting an inference result according to the above-mentioned fuzzy rules.

The same figure (A), CB) shows the two variables (X, , ,
The membership function corresponding to XZ) is shown, and the same figure (C
) represents the membership function corresponding to the consequent. Here, two membership functions for the antecedent part are shown, but as the types of variables in the antecedent part increase, the number of membership functions increases accordingly. In each figure, the horizontal axis represents the value of the variable, and the vertical axis represents the membership value (degree of belonging).

Now, the value of the variable x in the first item of the antecedent part is X.

Assuming that, the degree of affiliation at that time is 0.5 (see (A) in the same figure). Furthermore, if the value of the variable x2 in the second item of the antecedent part is x2°, then the degree of affiliation is 0.3 (see (B) in the same figure). In such a case, the fuzzy calculation unit takes the smallest value among the degrees of membership. That is, in the above example, a degree of affiliation of 0.3 is selected. Next, the membership function corresponding to Z is cut off at the above membership degree of 0.3, and the center of gravity position y" of the lower trapezoidal part S is obtained. Then, one rule is to output this yo as the inference result. The above inference is made for , but generally multiple rules are set. In this case, the 10th rule is set for each rule.
The inference result shown in Figure (C) is output. In this case, the trapezoid parts output for each rule are logically summed, and the center of gravity y of the logically summed part (hatched area in FIG. 1θ (D)) is outputted as the determined value of the inference.

In the above inference method, the logical product operation rule of the degree of belonging to the antecedent part (operation to select the smaller degree of membership) and the logical sum operation rule of the trapezoidal part to the consequent part are
It is called the ax rule.

Now, FIG. 1 shows the configuration of the invention. In the figure, a TV camera 1 captures an image of a soldered part of an electronic component, and an image memory 2 stores the image information. This TV camera 1 and image memory 2 correspond to image information generating means 3 according to the present invention. The feature amount extracting means 4 extracts the feature amount of the soldered portion of the electronic component from the obtained image information. The fuzzy inference means 5 uses the extracted feature quantity as a variable and outputs an inference result regarding soldering quality as an evaluation quantity of soldering quality.

The determining means 7 outputs the final soldering quality determination result in accordance with this evaluation amount. The membership function setting means 6 determines a membership function to be used by the fuzzy inference means 5, for example, in order to increase the determination rate by suppressing the proportion of evaluation quantities that belong to a so-called gray area that is neither good nor bad.

(flEmbodiment FIG. 2 is an external perspective view showing the appearance of a portion to be inspected by a circuit board solder surface inspection apparatus according to an embodiment of the present invention.

A chip component is soldered on a circuit board, and two slit lights a15a and 1 emit slit light in the long axis direction of the chip component through the soldered part (solder fillet).
5b is placed. Further, a slit light source 15c is arranged which is orthogonal to the slit light in the long axis direction and irradiates the slit light in the short axis direction of the chip component. Furthermore, a TV that images the projection pattern of these slit lights onto the chip components and their vicinity from diagonally above.
A camera 1 is arranged. Three slit light sources 15a
~15c and the TV camera 1 are fixed, and the shape of the soldered part is read by moving the XY child table on which the circuit board is placed and placing the part to be inspected at the intersection of the long axis slit and the short axis slit. .

FIG. 3 is a block diagram of the control section of the circuit board solder surface inspection apparatus.

The CPU 10 executes a program written in advance in the ROMII to perform overall control of the entire system. RAM12
has an area for storing a profile image of the soldering part taken by the TV camera 1 and other working areas. The I10 interface circuit 13 includes driver circuits 14a to 14 that drive the slit light sources 15a to 15c.
4c is connected. The CPU10 controls these driver circuits via the I10 interface circuit 13 to emit the slit light at the required timing. Note that an inspection start signal is input to the I10 interface circuit 13, and based on this signal, the CPU 10 inspects the electronic components on the circuit board determined by the current XY child table position. When the inspection is completed, an inspection completion signal is outputted to a circuit board inspection device (not shown) via the I10 interface circuit 13. This allows the circuit board inspection device to determine whether, for example, the next electronic component to be inspected is T.
The XY child table is driven so as to be within the imaging area of the V camera. A CCD area sensor 16 and a CCD controller 17 that controls the CCD area sensor 16 and generates a video signal are provided in the TV camera 1. The video signal controller 18 is a circuit that binarizes the output signal of the TV camera 1 at a predetermined threshold value and performs serial/parallel conversion.The CPU 10 provides a timing signal to the video signal controller 18, and Load the parallel-converted image data and store it in RAM1.
sequentially written into the image data storage area in 2. D/A
Converters 20 and 21 convert the solder fillet length and height data output from the corresponding interface circuit 19 into analog signals and provide the analog signals to the fuzzy inference section 5. The fuzzy inference unit 5 uses these two signals 1. A fuzzy calculation is performed using h as a variable, and a final value e representing the evaluation amount of soldering quality is output. A/D converter 22 converts this into a digital code. CPU10 reads the value of this evaluation amount via I10 interface circuit 19.

FIG. 4 shows the shape of the soldered portion of the chip component. As shown in the figure, the height of the solder fillet in contact with the electrode of the chip component is defined as the solder height, and the length of the part in contact with the land on the circuit board is defined as the solder length.These solder height and solder length are determined by the characteristics of the soldered part. @Extract as a legacy.

FIG. 6 is a flowchart showing the processing procedure of CPUIO. First, image data is fetched from the video signal controller 18, and the image data is fetched into a predetermined area in the RAM 12 (nl). Next, the characteristics of the soldered portion of the electronic component, ie, the solder height and solder length of the solder fillet, are extracted from the captured image data (n2). Once the feature quantity has been extracted, the value is output to the fuzzy inference section 5, and the evaluation quantity, which is a definite value, is read from the fuzzy inference section 5 (n3-=n4-n5). Next, the evaluation amount is judged in three stages: good soldering, bad soldering, neither good nor bad, according to a predetermined range, and the results are output (n6
-=n7). Based on the determination result from the circuit board solder surface inspection device, the circuit board inspection device moves the XY child table to inspect the soldered portion of the next electronic component if the soldering is good. If the soldering is defective or neither good nor bad, the circuit board is removed from the production line. Note that if no electronic component is mounted, the answer is NO at step n3, and a code indicating this is output at step n8.

The method for extracting the feature amount of the soldered portion in step n2 will be described below. FIG. 7 is a diagram explaining the processing procedure of the CPU, and FIG. 5 is a diagram explaining the processing process. First, as shown in Figure 5 (A), a virtual horizontally long rectangular window is scanned vertically within the image data storage area, and the long axis is the center of gravity of the window when the total brightness within the window is greatest. Find it as the slit position. Next, as shown in the same figure (B), a minute window is scanned along this slit from the center to the outside, and the center of gravity of the window when the total brightness within the window becomes dark is determined by the length of the chip component. Let it be the end point of the direction. At this time, the influence of noise can be eliminated by detecting when the total brightness within the window has become darker several times in succession and finding the center of gravity of the last bright window. Next, as shown in FIG. 4C, a horizontally elongated window is drawn downward from the end point to find the center of gravity. Next, the same window is drawn below the center of gravity in the horizontal direction to find the center of gravity.

By repeating this process up to the substrate surface, the profile of the solder fillet is detected. In addition, Fig. 5 (A) to (C
), the short-axis slit pattern (see FIG. 2) is omitted. The pattern of short axis slits is used for inspecting the mounting position of parts or finely adjusting the position of the XY child table.

After determining the profiles of the long axis slit and solder fillet in this manner, the length and height from the end point position of the solder fillet on the substrate to the end point of the chip component are determined. In addition,
The height is determined as a ratio to the distance between the long axis slits (electrode height) from the substrate surface.

In this embodiment, the inference rule for expressing the evaluation amount of soldering quality is as follows.

-Inference rule- (1) If the solder length is above normal and the solder height is above normal, then the soldering is good.

(2) If the solder length is short and the solder height is above normal, the soldering is neither good nor bad.

(3) If the solder length is longer than normal and the solder height is low, the soldering is almost good.

(4) If the solder length is short and the solder height is low, the soldering is almost always poor.

When the entire rules including the rules shown above are expressed as a matrix, it becomes as shown in FIG. 8. In FIG. 8, each symbol regarding the solder length means ZR: normal or more NS: short NL: none.

Each symbol regarding solder height is ZR: Above normal NS: Low NL: None It means.

The contents of the matrix are soldering evaluation quantities, and each symbol means PL: Good PS: Very Good ZR: Intermediate (neither good nor bad) NS: Mostly Bad NL: Bad There is. Note that PL, PS, ZR, NS, and NI expressing the above-mentioned ambiguous linguistic values are called labels. Ambiguous linguistic values, i.e. the labels PL-NL above
The membership functions expressing
). FIG. 3(A) shows the membership function of solder length, and FIG. 2(B) shows the membership function of solder height. In addition, FIG. 6C shows membership functions for evaluation quantities for determining the quality of soldering.

FIG. 10 is a block diagram of the fuzzy inference section 5.

As mentioned above, the fuzzy inference section 5 includes the fuzzy operation section 20
and a defuzzify section 21. In order to output the inference result Xi for each rule according to each inference rule shown in FIG. It has a membership function generator for output. Since each fuzzy calculation unit is provided for each rule, a total of nine units, 20a to 20i, are provided, and the inference results Xi of each fuzzy calculation unit are processed in parallel by the defuzzify unit 21.
is output to.

The fuzzy calculation section has a configuration as shown in (1) FIG. 1(A). Note that this figure shows a configuration diagram of the fuzzy calculation section 20a.

Three general purpose membership function generators 3 as shown
0 to 32, and a label ZR1 corresponding to the solder length l, a label ZR1 corresponding to the solder height 1), and a label PL corresponding to the evaluation me are input to each membership function generator. When this label is input, a general-purpose membership function generator generates a membership function corresponding to the label. That is, in the fuzzy operation unit 20a shown in FIG. 1(A), the membership function shown on the rightmost side of FIG. 9(A) is generated in the membership function generator 30, and the membership function generator 31
In this case, the rightmost membership function shown in FIG. 9(B) is generated. Further, the membership function generator 32 generates the rightmost membership function in FIG. 9(C). Output of membership function generator 30.31,
In other words, the degree of belonging of each term in the antecedent part is output to the antecedent part logical product 33, and here m1n of the mini-max rule mentioned above is
The smaller degree of affiliation is selected by the i-rule.

The result is sent to the consequent AND circuit 34.

The consequent AND circuit 34 applies the inference result from the antecedent AND circuit 33 to the membership function output from the membership function generator 32, as shown in FIG. 13(C).
Perform the head truncation (perform logical product) as shown in , and output the trapezoidal part as the inference result.

The defuzzifier 21 has the configuration shown in FIG. 1(B). As shown in the figure, the defuzzifier 21 includes an OR circuit 40 and a definite value calculation circuit 41. The logical sum circuit 40 is a part that calculates the max rule of the mini-wax rule, and it logically adds the trapezoidal outputs (inference results) from each of the nine fuzzy calculation units, and the result is shown by hatching in FIG. 13(D). form a region like this. The fixed value calculation circuit 41 determines the center of gravity position from this area and outputs it as an evaluation amount e of soldering quality.

Next, the specific operation of the above device will be explained.

For example, as shown in Figure 9(A), the solder length is 162μ.
Suppose that m. In this case, the label for the solder length l (hereinafter expressed as 1 (00)) 1 (Z
Only the membership degrees of membership functions corresponding to R) and l(NS) are displayed. The membership degree of the membership function corresponding to label β(NL) is O. That is, J:ZR=0.55 1:N5=0.15.

In addition, as shown in the same figure (B), the solder height is 40 mm higher than the electrode height.
%, only the degrees of membership of membership functions corresponding to labels h (ZR) and h (NS) are displayed. That is, h:ZR=0.45 h:N5=0.2.

Looking at the above membership degrees, we can see that only operations corresponding to rules have a label of ZR or NS in the antecedent part regarding solder length l, and a label of ZR or NS in the antecedent part regarding solder height h. It will be done.

Next, let's look at the operation of the antecedent AND consequent circuit and the consequent AND circuit.

In the antecedent logical product circuit, m1n of the mini-max rule
Take the smallest value for each rule according to the i-rule. Then, in the consequent logical AND circuit, the membership function of the label corresponding to the consequent is truncated.

From the table shown in Fig. 8, in the example shown in Fig. 9, in the antecedent part AND circuit of the fuzzy operation part whose antecedent part labels are l (ZR) and h (ZR), the label 1 (ZR) is Since the degree of belonging is 0.55 and the degree of belonging to Rahe Jl/h (Z R) is 0.45, the smaller value of 0.45 is selected. In the antecedent logical product circuit of the fuzzy operation part whose antecedent labels are 1 (ZR) and h (NS), the label 1
Since the affiliation degree of (ZR) is 0.55 and the affiliation degree of h (NS) is 0.2, the smaller value of 0.2 is selected. Also, the labels of the antecedent part are / (NS) and h (ZR
), the degree of membership of label 1 (NS) is 0.15 and the degree of membership of h (ZR) is 0.45, so the smaller value of them is 0.
15 is selected. Furthermore, the label of the antecedent part is ffi (
In the antecedent logical product circuit of the fuzzy operation unit, which is NS) and h (NS), the degree of membership of label J (NS) is 0,15.
, h(NS) is 0.2, so the smaller value of 0.15 is selected.

Next, in the consequent logical AND circuit, the membership function of the label corresponding to the consequent is truncated.

By performing the above operation in the fuzzy operation section, the defuzzify section 21 is finally
) as shown in the label e (NS)e (ZR)
, e (PS), and e (PL), a trapezoidal part (hunting part) obtained by truncating the membership functions corresponding to each of them is output as an inference result. Defuzzify section 21
The logical sum circuit 40 logically sums these four trapezoidal parts.

That is, the max rule of the mini-max rule is executed here. Next, a calculation for determining the center of gravity of the superimposed trapezoidal portions is performed by the definite value calculation circuit 41, and the calculated center of gravity position e1 is outputted as an evaluation quantity of soldering quality. In this case, the evaluation amount e is within the good judgment area indicated by - in FIG. 9(C).
1 exists, this soldered part is determined to be "good".

In the above embodiment, three types of membership functions were set to represent the height and degree of length of the solder fillet, but for example, as shown in FIGS. It is also possible to divide the degree of sag into four types. Further, as information representing the characteristic t'fxM of the soldered portion, the degree of gloss of the solder can also be detected by secondary reflection. Secondary reflection is a phenomenon in which the reflected light of the irradiated slit light on the solder surface is reflected by a TV camera lens, etc., and irradiates the solder surface again, and the brighter this is, the better the soldering condition is likely to be. Therefore, in addition to incorporating this secondary reflection into the inference rule, it is also possible to use a membership function related to the brightness of the secondary reflection as shown in FIG. 12(C). However, since secondary reflections due to incompletely soldered parts also occur, it is necessary to take this into account in the rules.

Further, in the embodiment, a specific area of the RAM is used as an image information storage area, but a dedicated image memory for capturing image data by DMA may be provided.

In this case, since the image reading speed is improved, the time required for the inspection can be shortened.

(g> Effects of the Invention As described above, according to the present invention, the evaluation amount for determining the quality of soldering of electronic components by fuzzy inference from the characteristics (■) of the soldered parts of electronic components mounted on a circuit board This makes it easy to judge whether the soldering is good or bad.In addition, when judging the soldering quality in three stages: good, bad, and neither good nor bad, it is possible to easily judge whether the soldering is good or bad. By observing the characteristics of the determined soldering condition of the circuit board and modifying the shape and position of the membership function used for fuzzy inference, we can lower the proportion of evaluation quantities that are neither good nor bad and increase the determination rate. It becomes possible to go.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of the present invention. Second
The figure is an external perspective view showing the appearance of a part to be inspected by a circuit board solder surface inspection apparatus according to an embodiment of the present invention, and FIG. 3 is a block diagram of a control section of the apparatus. FIG. 4 is a diagram showing the shape of the soldered portion. FIGS. 5(A) to 5(C) are diagrams for explaining a method for extracting a special quantity of a soldered portion. FIG. 6 and FIG. 7 are flowchart 1 showing the processing procedure of the control section. FIG. 8 is a diagram showing inference rules set in the embodiment. FIGS. 9A to 9C are diagrams showing membership functions. Figure 1O is a configuration diagram of the fuzzy inference section, Figure 1) (A
) and (B) are block diagrams of a fuzzy operation section and a defuzzify section, respectively. FIGS. 12A to 12C are diagrams showing member zip functions according to other embodiments. Further, FIGS. 13(A) to 13(D) are diagrams for explaining the operation of the fuzzy inference section. ■ -TV camera, one characteristic 1″&quantity extraction means, fuzzy inference means (fuzzy inference section)

Claims (1)

[Claims] (1) means for capturing an image of a soldered part of an electronic component to generate image information; means for extracting feature quantities of the soldered part of an electronic component from this image information; A solder surface inspection device for a circuit board, characterized in that it is provided with fuzzy inference means for obtaining an evaluation quantity for determining the quality of soldering of a component.