JP2857705B2 - Abnormal product detection method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a method for detecting an abnormal product in a manufacturing process in which an allowable range of a management item varies for each production unit.

As [Prior Art] PGA (P in G rid A rray ) for lead such applications, clad pin is welded to the head of the lead pin bonding material such as silver have been utilized, FIG. 7 is clad pin It is an enlarged plan view. The clad pin 10 has a thin cylindrical shaft 11
and a joining material 12 made of silver on the bottom surface of the head 11b of the lead pin 11 composed of a and a disk-shaped head 11b having a larger diameter than the shaft portion 11a.
Are hemispherically welded. The process of manufacturing such a clad pin is as follows.

Prepare a tray in which a large number of holes having substantially the same shape as the lead pins are formed in a lattice shape, and insert the lead pins one by one into each hole of this tray, with the head positioned on the opening side of the hole, One silver particle (joining material) is placed on the head of each lead pin. Next, the tray storing a large number of sets of lead pins and silver grains is carried into a heating furnace and heated at a high temperature. Then, the silver grains are heated, and are hemispherically welded to the heads of the lead pins as shown in FIG. 7 by surface tension to become silver solder, thereby producing clad pins.

In this manufacturing process, the mass of all silver grains in one tray is substantially the same and falls within the range of the normal distribution, but the state of the normal distribution differs between the trays. (The shape is substantially the same, but the average value is different.) Then, a clad pin in which a required amount of a bonding material is welded is manufactured. Further, it is known that selection of non-defective products and abnormal products in each manufactured clad pin is performed based on whether or not the area value (management item) is within an allowable range based on the imaging area of the bonding material in each clad pin. .

[Problems to be solved by the invention]

A manufacturing standard is set in advance for a product, and a product that does not conform to the standard is regarded as a defective product. Therefore, an inspection method for determining a defect detection threshold using a tolerance of the standard is generally adopted. Therefore, when manufacturing and inspecting various products having different manufacturing standards in the same process, it is necessary to change the threshold value each time according to the product type.

On the other hand, when inspecting manufacturing control items that cannot be directly measured as described above, thresholds are obtained experimentally, and in order to prevent overdetection, consider the changes in instruments and determine the optimal threshold for each product type taking into account allowable values. Must be determined empirically.

Detect abnormal products in a system where multiple products are manufactured at once for each production unit, such as the manufacturing process of clad pins, and the permissible range of control items differs for each production unit. In such a case, the detection is often performed by over-detection in order to prevent the abnormal product from being leaked to the user.

Further, when the allowable range is set in consideration of the product variation between the production units, there is a problem that the allowable range is narrowed and the yield is deteriorated. In order to determine the optimum threshold value for a new variety, a huge amount of time and data sampling are required. Further, at the time of operation, a threshold value change frequently occurs to cope with various kinds. From this,
A method of registering a threshold having an appropriate tolerance value for each product type and selecting the same, performing an operation of correcting the failure as needed on a daily basis while checking the defect detection status, and testing the same product type as much as possible together We have been working on the operation. However, even among such operations, it is not possible to avoid outflow of defective products due to input errors, threshold failures due to fluctuations in measuring instruments, and the like, and the burden on operators when varieties change frequently is extremely large. In particular, it is necessary to pay attention to the fluctuation of the illuminance for those involving image processing.

The present invention has been made in view of such circumstances, and obtains a histogram of a product for each production unit, and detects a product that is out of the population of the histogram as an abnormal product. It is not necessary to take into account the need to improve yield, and even for new varieties, there is no need to conduct experiments to collect a large number of data to determine the threshold value, and furthermore, an allowable range is set for each production unit It is an object of the present invention to provide an abnormal product detection method that does not need to be performed. The “population” described above is not a statistical one (a group consisting of observable individuals: according to the Iwanami Dictionary of Physical and Chemical Sciences), but refers to a continuous cluster of histogram classes as described later.

[Means for solving the problem]

The abnormal product detection method according to the present invention is a method for manufacturing a plurality of products collectively for each production unit, and furthermore, for a manufacturing system in which the permissible range of the management items of the manufactured product varies for each production unit, the permissible range is outside the permissible range. In the method of detecting a certain abnormal product,
For each production unit of a manufactured product, classify the numerical values of the control items of each product to obtain a histogram thereof, obtain the range of the population of the histogram, and detect products out of the range as abnormal products. Features.

[Action]

In the abnormal product detection method of the present invention, a histogram of a product manufactured for each production unit is obtained. Next, each product is selected to determine whether or not it is within the population of the histogram, and a product outside the population is detected as an abnormal product. Then, an abnormal product is detected based on the histogram obtained for each product in each production unit, so that there is no need to consider product variations among the production units.

〔Example〕

Hereinafter, the present invention will be described with reference to the drawings showing the embodiments. In this embodiment, a case where an abnormal product in the clad pin is detected will be described.

FIG. 1 is a schematic view of an apparatus for carrying out the method of the present invention. In FIG. 1, reference numeral 13 denotes a tray carried out of a heating furnace after clad pins have been manufactured. The tray 13 has a large number of holes in a lattice shape, and each of the holes stores one manufactured clad pin 10. The tray 13 has a position controller 8
Is mounted on the inspection stage 7 which can be moved in the XY direction (the left and right direction and the front and back direction in FIG. 1) by the control of, and can be moved in the XY direction integrally with the inspection stage 7.

The obliquely above the tray 13 and the light source 3 is provided also above the tray 13, in a manner that its optical axis is perpendicular to the welding surface of the bonding material 12, for example, CCD (C harge C oupled D
evice) A camera 2 comprising a camera is provided. By the illumination from the light source 3, the surface images of the heads of the plurality of clad pins 10 are captured by the camera 2.

In FIG. 1, reference numeral 1 denotes an image processing apparatus for processing a surface image obtained by the camera 2;
A binarization processor 4 for binarizing the surface image obtained in step 2, a labeling processor 5 for labeling the binarized image from the binarization processor 4 for each clad pin, and a labeling process An area / position measuring device 6 for measuring the imaging area of the bonding material of each clad pin and the position of each clad pin based on the binarized image is provided.

The area / position measuring device 6 outputs the measured area information and position information to an abnormal product detection device 20 that detects an abnormal product based on the information. The abnormal product detection device 20 calculates a population of the histograms created by the histogram creator 21 and the histogram creator 21 based on the information from the area / position measuring device 6. Population calculator 22, a memory 23 for storing information from the area / position measuring device 6, and an abnormal product detector 24 for detecting an abnormal clad pin based on the population of the histogram and the contents stored in the memory 23. I do.

 Next, a specific procedure of the method of the present invention will be described.

First, the position controller 8 is driven to move the inspection stage 7 (tray 13). Light is emitted from the light source 3 to the plurality of clad pins 10 located below the camera 2, and the camera 2 images the heads of the plurality of clad pins 10.

The image obtained by the camera 2 is input to the image processing device 1 (binarization processor 4), and is binarized by the binarization processor 4 at a predetermined luminance level. The binarized image is
The labeling processor 5 labels a unique number for each clad pin. Based on the labeled binarized pixels, the area / position measuring device 6 measures the imaging area of the bonding material for each clad pin and the position (XY coordinates) of each clad pin. Then, the area / position measuring device 6 outputs the area information and the position information to the abnormal product detecting device 20 (the histogram creating device 21 and the memory 23).

The histogram creator 21 creates a histogram as shown in FIG. 2 at appropriate class intervals based on the area information of each clad pin and the total number thereof. Needless to say, the vertical axis in FIG. 2 represents the frequency. The population computing unit 22 computes both ends (a ₁ and a _{2 in} FIG. 3) of the population based on the histogram as shown in FIG. 3, and sends the region of the population to the abnormal product detector 24. Output. The population refers to a continuous cluster of histogram classes as shown in FIG.

FIGS. 4, 5, and 6 show a histogram generator 21 and a population calculator 2
3 is a flowchart showing the operation of No. 2, and these operations will be described below based on this flowchart. FIG. 4 is a main flowchart of the operation, and FIGS.
FIG. 6 is a flow chart showing the operation of smooth differentiation and population range calculation in FIG. 4, respectively.

A class interval for creating a histogram is set in the histogram creating unit 21 in advance, and the area / position detector 6 is set.
Each time area information is input from, it is determined which class this input value belongs to and added to the cumulative value (number of elements) of that class (S1). After performing this processing for all input values (S2), smooth differential processing is performed on the input values (S2
3).

For each class divided at a predetermined class interval, the i-th (n + for all classes) on both sides of each class (n) centered on each class (n)
The difference between the accumulated values in the (i-th and ni-th) ranges is determined (S3
2) The process of adding these (S33) is repeated a predetermined number of times. After repeating this, it is determined whether the obtained differential smoothed value is 0 (S34), and if it is 0, the accumulated value is set as the differential smoothed value of the class (S35). The above processing is performed for all ranges, and when it is determined that the processing has been performed for all classes (S30), the process returns to the main routine of FIG. 4 (S36), and the next population range calculation step ( Enter S4).

In the flowchart shown in FIG. 6, the left end of the population is searched first, and then the right end is searched. First, it is determined whether the cumulative value in a class is larger than a value that can be defined as a group (S41). If smaller, it is determined that the user has not entered the population, and the next class is searched. Next, the sign of the increase code is checked (check whether the differential smoothed value is positive) and the sign of the bottom width (the count value added when the accumulated value is smaller than the value defining the bottom) is checked (S42). ). Increment code> not 0 or bottom width>
If it is not 0, it is determined that it is not the left end, and it is determined whether the accumulated value is larger than the value defining the bottom (S4
3) If the accumulated value is small, the count of the bottom width is incremented (S44), and the next class is searched. When the increase image is> 0 and the bottom width is> 0,
The class is determined to be the left end of the population, and this position the left end a _1, clears the counting of the bottom width (S45). Next, the count of the group width is incremented. If the count is greater than 1 (S46), the search for the right end of the population is started (S47). And similarly to detect the right edge a ₂ populations.

The memory 23 records information from the area / position measuring device 6, and the recorded content is read out to the abnormal product detector 24 by a read signal from the abnormal product detector 24.

The abnormal product detector 24 reads the area information of each clad pin from the memory 23, determines whether or not the area value falls within the range of the population, determines that the clad pin included is a good product, and removes it. The clad pin (FIG. 3A) is determined to be abnormal.

In the present embodiment, the case of detecting an abnormal product in the clad pin has been described. However, it is possible to perform the same operation as in the present embodiment in another manufacturing system in which the allowable range of the control item varies for each production unit. Of course.

Also, the method of the present invention can be applied to the case where the conditions of a measurement system (for example, a camera, a light source, and the like in the above-described embodiment of the clad pin) that measures the control items for each production unit change.

〔The invention's effect〕

As described in detail above, in the method of the present invention, a product that is out of the population of the histogram is detected as an abnormal product, so that it is not necessary to consider the product variation in each production unit, and the yield can be improved.

Further, when the present invention is applied to a manufacturing system in which the measurement condition of the control item varies for each production unit, the abnormal product is not considered for each production unit without considering the measurement condition variation for each production unit. Can be detected.

Even for new varieties, there is no need to determine the threshold value by conducting experiments for collecting a huge amount of data.

Further, it is not necessary to set an allowable range corresponding to each production unit. In accordance with this, the present invention has excellent effects such that there is no input error for setting, and even if fluctuations in the measuring instrument and illuminance occur, there is no possibility that a large number of products will be rejected due to such fluctuations. Play.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of an apparatus for carrying out the method of the present invention, FIG. 2 is a schematic diagram of a histogram, FIG. 3 is a schematic diagram showing a population, and FIGS. 7 is an enlarged plan view of the clad pin. 1. Image processing device 2. Camera 3. Light source 4.
… Binarization processor, 5… labeling processor, 6… area / position measurement device, 10… cladding pin, 20… abnormal product detection device, 21… histogram generation device, 22… population calculation , 23 …… Memory, 24 …… Abnormal product detector

 ──────────────────────────────────────────────────続 き Continuation of front page (56) References JP-A-1-259246 (JP, A) JP-A-62-273407 (JP, A) JP-A-61-187637 (JP, A) JP-A-58-1983 97614 (JP, A) Japanese Utility Model Showa 60-104759 (JP, U) JP-B 53-40904 (JP, B2)

Claims (1)

(57) [Claims] In a manufacturing system in which a plurality of products are manufactured collectively for each production unit, and an allowable range of control items of the manufactured product varies for each production unit, an abnormal product outside the allowable range is detected. In the method, for each production unit of the manufactured product, the numerical value of the control item of each product is classified and the histogram is obtained, the range of the population of the histogram is obtained, and the product out of the range is regarded as an abnormal product. An abnormal product detection method characterized by detecting.