JP2010085166A - Prepreg defect inspection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a defect inspection method of a prepreg surface formed by impregnating a matrix resin into carbon fibers, especially, a defect inspection method using an optical means, capable of inspecting the prepreg surface aligned in one direction, with high accuracy and high reliability. <P>SOLUTION: A rectangular rate, the maximum diameter angle and the unevenness degree of a defect candidate domain are determined, and a crack defect is detected. Then, a threshold is calculated for each pixel row from an image wherein a crack defect detection domain is masked, and a fluff defect is detected by performing binarization by using the threshold. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a method for inspecting defects on a prepreg surface in which a carbon fiber is impregnated with a matrix resin, and in particular, for inspecting defects using optical means capable of inspecting a prepreg surface aligned in one direction with high accuracy and high reliability. Regarding the method.

  Carbon fiber reinforced plastics (hereinafter referred to as CFRP) are widely used in various fields due to their high specific strength and high specific modulus, and require high strength in aircraft, boats, automobiles, etc. The use is expanding to the intended use. In these moldings, a sheet-like molding material called a prepreg in which carbon fibers are aligned and impregnated with a thermosetting resin is used, which is laminated and then cured.

  In a prepreg used for CFRP, a slight defect may cause a failure starting point and a strength decrease when CFRP is used, and therefore, a high accuracy inspection level is required in the pass / fail judgment of these inspection standards.

  These inspections were generally performed by the human eye, but automated technologies aimed at saving labor, improving accuracy, and suppressing oversight of defects have been introduced, reducing inspection time and improving inspection accuracy. For this reason, various techniques have been proposed.

  For example, Patent Document 1 proposes a prepreg manufacturing method and apparatus capable of detecting a plurality of types of defects with a single camera. This defect inspection method according to Patent Document 1 irradiates a prepreg with light, detects scattered light from the prepreg by a light receiving means, and compares the signal level of the scattered reflected light detected by the light receiving means with a preset threshold value. Have been identified.

Further, Patent Document 2 proposes a defect identification device for a sheet-like surface, and extracts a plurality of feature amounts from regions that are candidates for defects of a detected image as well as a signal level, and sets a predetermined defect. The defect is identified by comparing with the feature quantity of the kind.
JP-A-9-225939 JP2004-109069

In the conventional defect inspection method, there is a problem that a normal part is erroneously detected as a defect, a defect whose shape is very similar is detected as another defect, or a defect is detected as several types of defects. there were. In particular, in the prepreg aligned in one direction, the unevenness of the luminance value along the running direction axis extends, and the occurrence of a problem that is erroneously detected as a crack defect or a fluff defect is remarkable. The present invention has been made in view of such circumstances, and an object thereof is to provide a defect inspection method for solving the above-described conventional problems and accurately identifying defects.

  In the prepreg aligned in one direction, the present inventors have found that the surface of the prepreg has uneven coating of the resin and unevenness of the surface. Therefore, even if the prepreg is irradiated with light from a single direction, the prepreg Depending on the position of the light, it is reflected (scattered) in different directions. If such scattered light is imaged by the light receiving means, the luminance distribution of the captured image will not be uniform even in the normal part, and the defective part will be different from the normal part. Even if the luminance value is relatively high, a study is made based on the idea that a part of the normal part may have the same signal level as when the reflected light from the defect is received. The present invention was conceived. That is, such a normal portion having a high luminance value signal level is generated because there is no regularity in the unevenness of coating of the resin and the unevenness of the surface of the prepreg, and is irregularly distributed. Recognizing that it may appear as a shape that closely resembles the shape of the defect, after detecting the part of the normal part where the signal level of the luminance value is high and the defect whose shape is very similar to each other, the information is detected first. The present inventors have found a defect inspection method for accurately identifying defects by detecting other defects in consideration of the above.

In order to solve the above problems, the present invention has any one of the following configurations. (1) The surface of the traveling prepreg is irradiated with light by an illumination unit, the surface of the prepreg is imaged by an imaging unit, an image captured by the imaging unit is stored in a storage unit, and an image stored in the storage unit is stored. A method for inspecting a prepreg surface to detect and identify defects by processing, wherein a first binarization processing step of reading an image of the prepreg surface stored in a storage means and binarizing with reference to a predetermined threshold value Then, the bright area (A) is extracted from the first image data obtained by the first binarization processing step, the rectangular ratio, the maximum diameter angle, and the unevenness degree of the bright area (A) are obtained, and the crack defect is eliminated. A crack defect identifying step for identifying, a mask processing step for excluding an area that is a crack defect identified by the crack defect identifying step from the image of the surface of the prepreg stored in the storage means, and a previous step Integrated value Vs [x, y] defined below based on the mask image obtained from the mask processing step (y is an element of the traveling direction axis, x is an element of the axis orthogonal to the traveling direction axis) and one pixel column A second binarization processing step for binarization based on each threshold Vt [x];
A fluff defect identifying step for extracting a light area (B) from the second image data obtained by the second binarization processing step, obtaining a feature amount of the light area (B), and identifying a fluff defect This is a method for inspecting the prepreg surface. (In this specification, the bright regions (bright region (A) and bright region (B)) obtained after the binarization processing step may be collectively referred to as defect candidate regions from the meaning)

(2) The illuminating means is the prepreg surface inspection method according to (1), wherein the illuminating means is provided so that an incident angle with respect to the prepreg surface is 20 to 40 degrees when the prepreg surface is used as a reference.
(3) The prepreg surface inspection method according to (1), wherein the first illuminating means is line-shaped illumination, and is provided so that a longitudinal axis of the line-shaped illumination is perpendicular to a traveling direction of the prepreg.
(4) The prepreg surface inspection method according to (1), wherein the imaging unit is provided so that an optical axis thereof is perpendicular to the prepreg surface.

In the present invention, the “prepreg” refers to a sheet-like material obtained by impregnating a carbon fiber aligned in one direction with a matrix resin.
In the present invention, “line illumination” refers to illumination having a region in which the light emitting portion extends linearly in a one-dimensional direction and the luminance distribution along the longitudinal axis is constant. The light emitting portion may be an elongated surface having a width as long as the length in the direction orthogonal to the longitudinal direction is within 10 mm and substantially constant in the longitudinal direction. Here, “substantially constant” means that the deviation of the width of each part is within ± 5 mm with respect to the average of the widths in the longitudinal direction. Such "line illumination" is specifically a straight tube fluorescent lamp, a plurality of optical fibers arranged in a line to guide light from a light source such as halogen or LED with a light guide, the front of the illumination means For example, the illumination is performed by providing a cylindrical lens, but it is preferable to use LED line illumination from the viewpoint of easy change of the light amount and cost and maintainability. In the present invention, the “longitudinal axis” of the line illumination is defined as the long axis of the light emitting surface.
In the present invention, the “optical axis” of the imaging means is defined as a perpendicular line from the center of the light receiving surface.
In the present invention, the “rectangular ratio” is defined as the ratio of the maximum diameter 32 and the minimum diameter 33 of a straight line passing through the center of gravity 31 of the defect candidate region 30 shown in FIG.
In the present invention, the “maximum diameter angle” is defined as the relative angle 34 of the axis perpendicular to the running direction and the longitudinal axis direction of the maximum diameter 32 of the defect candidate region shown in FIG.
In the present invention, the “degree of unevenness” is defined as the ratio of the envelope peripheral length 35 to the peripheral length 36 of the defect candidate region shown in FIG.
In the present invention, the “cracking defect” refers to a defect in which a gap is generated between the prepregs because the carbon fiber bundle is not sufficiently expanded. The “collapse defect” is recognized by the appearance of white due to the reflection of the release paper through the gap.
In the present invention, the “fluff defect” refers to a defect in which a part of the carbon fiber is cut and the fluff is in a state of being fluffy or a bundle of yarns is agglomerated. The “fluff defect” has the same hue as the normal part of the prepreg.
In the present invention, the “incident angle” is defined as an angle formed by the optical axis of light irradiated perpendicularly from the light emitting surface of the illumination means and the prepreg surface.

  According to the present invention, cracks and fluff can be accurately identified without erroneous detection.

  Hereinafter, preferred embodiments of the defect inspection method of the present invention will be described with reference to the drawings. FIG. 1 is a schematic block diagram showing the configuration of a defect inspection method according to an embodiment of the present invention (in the figure, “steps” such as “first binarization processing step” are omitted). . In the present embodiment, an image of a traveling prepreg is captured, and a defect detection process is executed based on the captured image. For this purpose, as shown in FIG. 1, the defect inspection method according to the present embodiment reads the illumination / imaging means 1 for generating image data, the storage means 2 for storing the captured image, and the stored image data to determine the predetermined data. Detected by a first binarization processing step 3 for binarization processing using the threshold value, a crack defect detection step 4 for detecting a crack defect by extracting a plurality of feature amounts, and a stored image reading first defect detection step A mask processing step 5 for masking the defective region, a second binarization processing step 6 for obtaining a threshold value for each pixel column, and a binarization process, and extracting at least one feature amount to detect a fluff defect Fuzz defect detection step 7. The threshold value for the first binarization processing step is the threshold value for obtaining the average brightness of the fluff and crack defects and not detecting the fluff defect as a bright area (A) defect candidate area during the binarization process step. Decide in advance. In a second binarization processing step 6 to be described later, if there is a portion with a long and high brightness on the prepreg travel direction axis, the threshold value of the pixel row increases. The crack defect is long in the running direction and has high brightness, so if you try to calculate the threshold value for each pixel row in the presence of the crack defect, the pixel column with cracks will have a high threshold value, and fluff and other defects Cannot be accurately identified. Therefore, prior to the second binarization processing step 6, the crack defect is detected in the crack defect detection step 4, and the crack defect region is masked in the mask processing step 5. Then, the threshold value of the pixel row can be calculated without being affected by the crack defect.

  FIG. 2 is a diagram showing the configuration of illumination / imaging means according to an embodiment of the present invention. The illuminating means 21 used in the present embodiment is a line-shaped illuminating device, and is provided so that the longitudinal axis thereof is orthogonal to the traveling direction of the prepreg (only the contour is shown in the figure but black). Further, the incident angle with respect to the prepreg surface is set to 20 to 40 degrees when the prepreg surface is used as a reference.

As the type of illumination means, a line illumination device as shown in the present embodiment is preferable. A line-shaped illuminating device is an illuminating device having a shape in which a light emitting portion extends linearly in a one-dimensional direction, and is an illuminating device having a region in which a luminance distribution along a longitudinal axis thereof is constant. In which a plurality of optical fibers are linearly arranged to guide the light from the light source with a light guide, and in which a cylindrical lens is provided in front of the illumination means to illuminate, etc. Can be mentioned. For example, there are straight tube fluorescent lamps, LED line lighting, condensing type line lighting, and the like. In the present invention, LED line illumination is preferable from the viewpoints of easy variable light quantity and cost and maintainability. The length of the light emitting part of such a line illumination device is not particularly limited as long as the entire width of the prepreg can be illuminated uniformly, but it is preferable to cover the prepreg width to the outside of 50 to 150 mm on one side. For example, if the width of the prepreg is 1000 mm, the length of the line illumination is set to 1200, and the illumination may be performed to the outside by 100 mm. When a line illumination device is not used as the illumination means, it is necessary to arrange a light source so that the amount of light on the prepreg surface in the camera imaging range is uniform in the prepreg width direction.
The imaging means 22 is a line sensor. A line sensor obtains data related to brightness or darkness or luminance of one line. Light receiving elements (pixels) such as photodiodes that receive light are linearly arranged in a line, and the received light is converted into an electrical signal. This means that data is converted and output. For example, there are a CCD line sensor, a CMOS line sensor, an X-ray line sensor, and the like. In the present invention, it is preferable to have a light receiving sensitivity in the visible light region and to reduce noise in order to detect minute defects. From this viewpoint, a CCD line sensor is preferable. The inspection object irradiated by the line illumination device is imaged, imaging data of the inspection object is generated, and transferred to the storage unit 2.

  Next, the crack defect detection step 4 extracts the rectangular ratio, the maximum diameter angle, and the unevenness degree of the bright area (A) from the binary image obtained as a result of the first binarization processing step 3 by the crack feature amount extraction process 8. Then, a crack defect is detected by comparing with each predetermined feature amount. The bright area (A) is a bright area after binarization processing using the above-described preset threshold value, and is an area suspected of being a defect detected in this step (defect candidate area). For example, if the luminance depth is 0 to 255, the luminance value after binarization is either 0 or 255, so that the region where the luminance value is 255 is the bright region and the region where 0 is the dark region. . In addition, as the reference values for the rectangular ratio, maximum diameter angle, and unevenness, the average values of the rectangular ratio, maximum diameter angle, and unevenness of cracks and fluff defects are obtained, and cracks and fluff defects are accurately identified. It is determined in advance so as to be a value that can be drawn. In addition, before obtaining the rectangular ratio, the maximum diameter angle, and the unevenness degree, the defect candidate area may be narrowed down by deleting a minute area by a spatial filtering process. Further, before obtaining the rectangular ratio, the maximum diameter angle, and the unevenness degree, the number of pixels of each defect candidate area may be obtained, and a comparison value may be set in advance to narrow down the defect candidate areas.

  The mask processing step 5 reads an image from the storage means 2 and masks the area identified as crack in the crack defect detection step 4. Specifically, this is a process of setting the brightness value of the crack defect area to zero.

  In the second binarization processing step 6, a threshold value for each pixel column is calculated from the image in which the crack defect area is masked in the mask processing step 5, and binarization processing is performed based on the threshold value. As used herein, one pixel row is defined as the Y axis as the running direction axis and the X axis as the axis perpendicular to the running direction axis, as shown in FIG. 6, and the elements are y and x, and the number of each element is X: 10 Y: 5, Taking the case where each pixel is V [x, y] as an example, it indicates one column of V [a, 1] to V [a, 5]. In calculating the threshold value of one pixel column, first, an integrated value Vs [x, y] is obtained. The integrated value Vs [x, y] is an integrated value for several pixels on the X axis centered on each pixel,

And the above equation is calculated for all pixels. p <x is p ≧ x, for example, when Vs [2,1] integrated value is calculated with p = 2, the integrated range is V [0,1] to V [4,1] Refers to a non-existent pixel [0,1]. Therefore, the condition of p <x is necessary. The threshold value of the pixel column is compared with the integrated value Vs and a predetermined comparison value Vc, and a calculation for adding or subtracting a constant is performed for one pixel column. The comparison value Vc is obtained experimentally according to the average luminance value of the normal part. The value of p appears in the difference in the bright area (B) after binarization as shown in FIG. When p = 0, pixels in one pixel column (hereinafter referred to as vertical stripes) in which pixels with high luminance occupy most of the column are not detected as bright regions (B), but other portions are detected. On the other hand, when p = 1, the vertical streak and the pixels of the one pixel column on both sides thereof are not detected as the bright region (B). The calculation of one pixel column will be described using FIG. 8 as an example. In FIG. 8, the threshold value of one pixel column of x = 3 is obtained under the conditions of p = 1, Vc = 75, and k = 1. First, an integrated value is obtained for each pixel V [3,1] to V [3,5]. next

Vr of x = 3 is obtained under the conditions of Taking Vr [3,1] as an example, since Vs [3,1] = 81 and Vc = 75, Vs> Vc and Vr [3,1] = Va. On the other hand, since Vs [3,4] is Vs [3,4] = 67, Vs <Vc and Vr [3,4] = − Va. After calculating Vr for each pixel,

The Vr of each pixel is integrated according to the following formula. Taking FIG. 8 as an example, since the total value of Vr of 5 pixels is 3Va and k = 1, Vt [3] = 3Va. When attention is paid to V [3, 2] in FIG. 8, the luminance value is extremely higher than other pixels. However, even if this value is 28 or 255, Vt [3] = 3Va, not the magnitude of the pixel luminance value, but how many pixels with a luminance value higher than Vc exist in the one pixel column. It can be seen that the threshold is determined. Therefore, even if there is a pixel having an extremely high luminance value, the threshold value of the column does not become an extremely high threshold value.

  The fluff defect detection step 7 uses at least one feature amount (for example, the number of pixels, length, width, etc.) using the bright region (B) as a defect candidate from the binarized image obtained by the second binarization processing step. The fluff defect is detected by comparing with a predetermined value. The predetermined value is, for example, a value such that if the feature quantity to be obtained is the number of pixels, if the number of pixels is equal to or greater than that value, it is regarded as a fuzz defect, and if it is less, it is not regarded as a defect. And experimentally obtained in advance according to the defect standard.

  Further, as a pre-process prior to the second binarization processing step, a process is performed in which an image is read from the storage means and the image subjected to the horizontal filtering process is added to the image in which the crack defect area is masked from the mask processing step 6. Is preferred.

[Example 1]
An embodiment according to the present invention will be described. The prepreg to be inspected is 1000 mm wide and travels at 7 m / min. As the line-shaped illumination, one high-luminance LED line illumination (HLND type manufactured by CCS) was used, and it was arranged so that light was incident on the entire surface of the prepreg. A line sensor is used as the imaging means. Specifically, two line sensor cameras (NED e7450D) having the performance of 7450 elements, driving frequency of 40 MHz, and maximum scanning cycle of 192 μs are installed in parallel. The resolution of the prepreg in the width direction of the prepreg was adjusted to be 0.1 mm / pixel or more. The illumination angle was set so that the incident angle with respect to the prepreg surface was 30 degrees, and the amount of illumination light was set so that the illuminance on the prepreg surface was 9000 lx. The line sensor is installed so that its optical axis is perpendicular to the prepreg surface. In addition, light shielding plates are attached to the top and side surfaces so that external light does not directly strike the prepreg surface of the camera imaging range.

  The image data captured by the camera is temporarily stored on an image processing board attached on a personal computer, converted into an image of 7450 × 1000 pixels on the board, and then read into a memory on the personal computer.

  The brightness depth was set to 255 gradations, the average brightness value of the plurality of fluff and crack defect images was obtained, and the threshold value of the first binarization processing step was set to 45. With this value, after binarization, a first binarization process was performed, and 75 bright areas (A) were detected as defect candidate areas. Obtain the rectangular ratio, maximum diameter angle, and unevenness of all bright areas (A), set the maximum diameter angle range to distinguish from crack defects to 75 to 115 degrees, set the rectangular ratio to 5 or more, and unevenness to 1.5 or less. The crack defect detection process was performed. The range of the maximum diameter angle is set to 90 ± 25 degrees because the maximum diameter angle of the crack defect is the cause of the occurrence, and it is almost horizontal to the running direction axis. The rectangularity and the degree of unevenness were determined by determining feature amounts of a plurality of defects that are not repelled by identification based on the maximum diameter angle. In the calculation of the threshold value used in the second binarization processing step, the threshold value is calculated for each pixel column with p = 1, Va = 1, Vc = 90, and k = 1. The number of pixels in the region (B) was obtained. Each value Va, k is set to 1 in advance, and Vc is (1 + 2p) × average luminance value of the normal part, and the value of p that minimizes the influence of the normal part having a high luminance value is experimental. Asked. Fluff defect detection processing was performed using the number of pixels of 500 or more as a fuzz defect. Note that the number of pixels identified as fluff defects was obtained in advance as the number of pixels of a plurality of fluff defects, and the value of the fluff defect having the smallest number of pixels was used. Table 1 shows a result of extracting a defect candidate region whose shape closely resembles a crack defect and a result of detecting a fuzz defect.

  In this example, labels 230, 400, and 507 satisfy the crack defect condition, and the crack defect could be detected without erroneously detecting a defect having a very similar shape and a normal part. Moreover, fuzz defects could be detected without erroneous detection.

[Example 2]
The defect processing is performed in the same state as in the first embodiment except that the prepreg is driven at 5 m / min and the camera capture setting is changed so that the length in the traveling direction of the prepreg in one captured image is the same as that in the first embodiment. Went. The results are shown in Table 2. In this example, Label 89, 287, and 412 satisfy the crack defect condition, and the crack defect could be detected without erroneously detecting the defect having a very similar shape and the normal part. Moreover, fuzz defects could be detected without erroneous detection.

[Example 3]
The defect processing is performed in the same state as in the first embodiment except that the prepreg is run at 3 m / min and the camera capture setting is changed so that the length in the running direction of the prepreg in one captured image is the same as that in the first embodiment. Went. The results are shown in Table 3. In this example, Label 65, 87, 408 satisfies the crack defect condition, and the crack defect can be detected without erroneously detecting the defect having a very similar shape and the normal part. Moreover, fuzz defects could be detected without erroneous detection.

  The above-described defect inspection method of the present invention is preferably applied to a prepreg in which a carbon fiber aligned in one direction is impregnated with a matrix resin. As an application object, a black fiber sheet-like material aligned in one direction is used. As long as it is, it is not limited to this.

It is a schematic block diagram which shows the structure of the defect inspection method by one Embodiment of this invention. It is a figure showing the structure of the illumination and imaging means of one Embodiment of this invention. It is a figure for demonstrating the rectangular rate of a defect candidate area | region. It is a figure for demonstrating the maximum diameter angle of a defect candidate area | region. It is a figure for demonstrating the unevenness | corrugation degree of a defect candidate area | region. It is a figure for demonstrating the definition of 1 pixel row at the time of calculating a threshold value in the 2nd binarization process step. It is a figure for demonstrating the difference of the defect candidate area | region by the value of p in a 2nd binarization process step. It is a figure for demonstrating the calculation of the threshold value of 1 pixel row | line | column in a 2nd binarization process step.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Illumination / imaging means 2 Storage means 3 First binarization processing step 4 Crack defect detection step 5 Mask processing step 6 Second binarization processing step 7 Fluff defect detection step 8 Crack feature extraction process 9 Fluff feature Extraction Processing 10 Storage / Calculation Device 20 Prepreg 21 Illumination 22 Camera 30 Defect Candidate Area 31 Defect Candidate Area Center of Gravity 32 Defect Candidate Area Maximum Diameter 33 Defect Candidate Area Minimum Diameter 34 Defect Candidate Area Maximum Diameter Angle 35 Defect Candidate Area Envelope perimeter 36 Perimeter length of defect candidate area 40 10 × 5 image 41 Y-axis element number 42 X-axis element number 43 1 pixel array:

Claims (4)

The surface of the traveling prepreg is illuminated by illumination means, the surface of the prepreg is imaged by the imaging means, the image captured by the imaging means is stored in the storage means, and the image stored in the storage means is processed. A method for inspecting a prepreg surface for detecting and identifying a defect, wherein a first binarization step and a first binarization processing step for reading out an image of a prepreg surface stored in a storage means and binarizing with reference to a predetermined threshold From the first image data obtained by the binarization processing step, a bright area (A) is extracted, the rectangular ratio, the maximum diameter angle, and the unevenness degree of the bright area (A) are obtained to identify the crack defect. A masking step for excluding an area that is a crack defect identified by the crack defect identifying step from an image of the surface of the prepreg stored in the storage means and the mask; The integrated value Vs [x, y] defined below based on the mask image obtained from the processing step (y is an element of the traveling direction axis, x is an element of the axis orthogonal to the traveling direction axis) and 1 A second binarization processing step for binarization based on a threshold value Vt [x] for each pixel column;
From the second image data obtained by the second binarization processing step, a bright region (B) is extracted, a feature amount of the bright region (B) is obtained, and a fluff defect identifying step for identifying a fluff defect is performed. A method for inspecting a prepreg surface, comprising:

  2. The prepreg surface inspection method according to claim 1, wherein the illumination unit is provided so that an incident angle with respect to the prepreg surface is 20 to 40 degrees when the prepreg surface is used as a reference.   The prepreg surface inspection method according to claim 1, wherein the first-stage illumination means is line-shaped illumination, and is provided such that a longitudinal axis of the line-shaped illumination is in a direction perpendicular to the traveling direction of the prepreg.   The prepreg surface inspection method according to claim 1, wherein the imaging unit is provided such that an optical axis thereof is perpendicular to the prepreg surface.

Images