JP4509833B2 - Surface defect inspection method for aluminum alloy bar - Google Patents

Description

  The present invention relates to a method for inspecting surface defects of aluminum alloy bars such as aluminum alloy cast bars for forgings and aluminum alloy forged bars.

  Below, the case of the aluminum alloy casting rod for forging materials among aluminum alloy rod materials is demonstrated as an example. In recent years, aluminum alloy forged parts represented by suspension parts (hereinafter also referred to as suspension parts) have been used to reduce the weight of the vehicle body from the viewpoint of improving the fuel efficiency of automobiles. Such an aluminum alloy lower arm or upper arm suspension part is manufactured by hot forging a cast rod made of aluminum alloy. Along with the increasing adoption of these aluminum alloy suspension parts due to the weight reduction of the automobile body, a large production line for hot forging aluminum alloy cast bars will produce hundreds of thousands of suspension parts per month.

  In large-scale production lines for these suspension parts, a considerable part is automated to increase production efficiency. For example, such a large production line is disclosed in Patent Document 1 and the like (the entire configuration will be described later in detail), and aluminum alloy casting is performed by paralleling a plurality of relatively small-diameter casting bars. These casting rods are continuously cut into short pieces, subjected to external cutting (chamfering), and then subjected to homogenization heat treatment and hot forging. In this case, typically, the process from casting an aluminum alloy casting rod to before homogenizing heat treatment is automated.

  In such a large production line for automobile suspension parts, it is possible to inspect and sort defects on the surface and inside of the cast bar before forging as well as the product after forging for quality assurance of the suspension parts that are safety parts. Inevitable.

  Usually, inspection of internal defects of cast bars can be performed by ultrasonic flaw detection as a non-destructive inspection method, and this ultrasonic flaw detection has conventionally been automated (mechanized) in production lines in various fields. Has been done. However, surface defects (scratches) of an aluminum alloy cast rod cannot be inspected by ultrasonic inspection by a conventional method. For this reason, another test | inspection means for test | inspecting the surface defect of a cast bar is needed.

  In recent years, image analysis of a product surface using a CCD camera has been used as a non-destructive inspection means for the product surface. For example, an example using a CCD camera is known in order to eliminate a counting error in the number of round bars due to adhesion of ground iron powder in the end grinding line of round bars (see Patent Document 2). In this example, the whole end grinder of a round bar steel is photographed with a CCD camera, the color of the round bar steel in the image is stored in the image analysis device, and the color corresponding to the round bar steel counts the specified number in the image. The presence of a round steel bar is identified based on whether or not it exists within a specified area.

  In addition, as a defect inspection of the product surface, it is known that in a pickling line of a steel plate, a surface flaw accompanied with an oxide scale remaining on the surface of the steel plate is discriminated with a CCD camera and sent to a flaw removing device for processing ( (See Patent Document 3). In this example, in a conventional wrinkle inspection device that combines a CCD camera and a high-speed image analysis device, the detection accuracy of wrinkles is not sufficient, and if the judgment criteria are strict so as to eliminate oversight of wrinkles, the surface that is not originally required to be removed It has been pointed out that there is a problem that it is determined that the pattern should be removed.

For this reason, in Patent Document 3, in order to reliably detect only harmful surface defects remaining as defects after cold rolling without overlooking at the pickling stage, the resolution in the width direction and longitudinal direction of the coil is finer than 1000 μm. It is disclosed that it is preferable to use a high-resolution visible light CCD camera. In addition to the primary determination, this determination is performed not only for the primary determination but also for the image data of the part that is determined to be suspected to be removed by the primary determination device. It discloses that a wrinkle inspection device having a next determination device is preferable.
JP 2003-19533 A (claim, paragraphs 0016 to 0030, FIG. 1 and FIG. 2) JP 2003-220532 A (claim, FIG. 1) JP 2004-160511 A (claim, paragraph 0018, FIG. 2)

  In the above-mentioned Patent Document 1, only ultrasonic flaw detection is disclosed as a defect inspection of a cast bar, and the defect inspection on the surface of the cast bar described above is insufficient only by ultrasonic flaw detection. For this reason, it is effective to use a CCD camera for the defect inspection of the cast bar surface.

  However, in the defect inspection process of the casting rod surface in the forging line of the aluminum alloy casting rod, even if the CCD camera is used, minute irregularities, moisture, oil stains, etc. on the casting rod surface are removed from the actual casting rod. It is necessary to distinguish from surface defects (surface defects) of about several hundred μm to several tens of mm.

  In other words, the cast bar having these actual surface defects is either remelted as scrap as a defective product, or the surface is scratched (re-care). On the other hand, as described above, the surface of the cast rod that has fine irregularities due to external machining or has various stains such as moisture and oil stains on it has no problem in terms of quality. Need to be sent to the next heat treatment step. If such a good product is identified as an actual surface defect of the casting rod, a large number of casting rods with minute irregularities due to external machining or with various stains such as moisture and oil stains attached As a result, the yield and production efficiency of an automated large-scale production line for automobile suspension parts are significantly hindered.

  FIG. 9 shows the irregularities on the surface of the cast bar and the stains on the surface of the cast bar that are imparted by the external machining. In FIG. 9, FIG. 9 (e) shows minute, shallow and regular irregularities on the surface of the cast rod provided by the external machining. Further, FIG. 9F shows water droplets adhering to the casting rod surface during the process or prior to defect inspection, or deposits such as fine powder chips or cutting oil remaining without being removed by the cleaning. In the present invention, these are collectively referred to as “dirt” on the surface of the cast rod.

The surface defects of the actual cast bar are also shown in FIGS. The actual surface defects of the cast bar are various kinds of surface defects as shown below.
FIG. 9A: Entrainment (concave surface wrinkles with oxides and inclusions involved during casting).
FIG. 9 (b): Chigile (concave surface flaw with the aluminum surface peeled off during casting).
FIG. 9 (c): Surface defects associated with casting such as tension or blistering (deep scratches on the aluminum surface during casting, or concave surface flaws caused by gas entrained on the aluminum surface during casting).
FIG. 9 (d): A processing rod attached at the time of the external cutting.

  On the other hand, in Patent Document 3, a high-resolution visible light CCD camera having a resolution smaller than 1000 μm is used in order to improve the detection accuracy of surface defects accompanied with oxide scale remaining on the steel plate surface.

  However, in the defect inspection of the aluminum alloy cast bar surface, when a high-resolution visible light CCD camera with finer resolution as in Patent Document 3 is used, the minute surface Therefore, there arises a problem that regular irregularities and dirt on the surface of the cast bar are distinguished from actual surface defects of the cast bar. That is, it is pointed out in the above-mentioned Patent Document 3 that the criteria should be tightened so as not to overlook the wrinkles, and the surface patterns in FIGS. 6 (e) and (f) that do not need to be removed should be removed. There arises a problem that it is determined to be 疵.

  The present invention has been made to solve such a problem, and in the forging line of an aluminum alloy casting rod, the surface of the actual casting rod is subjected to minute irregularities on the surface of the casting rod, various stains, and the like. It is an object of the present invention to provide a method for inspecting a surface defect of a cast bar or other aluminum alloy bar in an aluminum alloy forged material manufacturing process that can be identified as a defect (surface defect).

In order to achieve this object, the gist of the surface defect inspection method of the present invention is a method for inspecting a surface defect of a manufactured aluminum alloy bar material. the brightness measurements carried out by the CCD camera, when the image analysis by the CCD camera, in a group of adjacent pixels, then obtained each average luminance values of two pixels adjacent to each other from the measured luminance of each pixel Further, the difference between the luminance average value of two adjacent pixels and the luminance average value of a pixel adjacent in the opposite direction to one of the two adjacent pixels (hereinafter referred to as “difference value of luminance average value”). The difference between the luminance average values is determined to be larger than a set value, and the occurrence of two or more consecutive occurrences in the adjacent group of pixels is identified as a surface defect, while the luminance average value difference If the value is less than all the set value, or that the difference value between the luminance average value that is greater than the set value identified as good product which does not occur in succession in a group of adjacent pixels said.

  In the present invention, as described above, when analyzing the image of the luminance measurement result by the CCD camera, the surface defect product and the non-defective product, that is, the surface defect of the bar, Identify surface stains and patterns.

  Usually, the image analysis of the luminance measurement result by the CCD camera used for detecting this kind of surface defect is identification of the surface defect based on the color signal level and the number of pixels of each pixel based on the binarization process.

  However, in such a commonly used image analysis, the irregularities on the surface of the cast bar and the dirt on the surface of the cast bar (hereinafter, also simply referred to as surface dirt, pattern, or non-defective product) due to the above-mentioned external cutting are used to determine the actual condition of the cast bar. It is impossible to distinguish or discriminate them from the surface defects (hereinafter also simply referred to as surface defects or defective products).

  On the other hand, according to the identification based on the luminance average value between adjacent pixels or the analysis based on the luminance average value difference (decrease rate) between adjacent pixels according to the present invention, Thus, it is possible to accurately identify the surface defects of the bar and the stains and patterns on the bar surface.

Embodiments of the present invention will be described below in detail with reference to the drawings.
First, a large-scale production line (forging line of an aluminum alloy casting rod) of forged aluminum alloy forged parts (suspension parts and the like) will be described with reference to FIGS. FIG. 7 is an overall configuration diagram of the manufacturing line, and FIG. 8 is a flowchart showing manufacturing steps of the aluminum alloy forged part. Note that these are disclosed in Patent Document 1.

  As shown in FIG. 7, the manufacturing apparatus 1 for forged aluminum alloy parts (hereinafter also referred to as forged parts) is a melting line for melting aluminum bullion and necessary additive elements as a line for producing an aluminum alloy cast bar. It comprises a furnace 2, a holding furnace 3 for finely adjusting the alloy component composition, a degassing apparatus 4 for performing degassing, and a continuous casting machine 5 for continuously casting billet B from the degassed molten metal. .

  Here, the continuous casting machine 5 includes a tundish 52 and a mold 53, and has a configuration in which the molten metal after degassing is poured from the trough 51 to the tundish 52. The continuous casting machine 5 introduces the molten metal poured into the tundish 52 into the mold 53 and continuously casts the billet B.

  Next, the aluminum alloy cast bar (billet) B is traveled and cut into a long cast bar L by a travel cutting machine 6, and the long cast bar L is further cut into a plurality of short cast bars S by a short cut machine 7. Short cut.

  The traveling cutting machine 6 has a cutting portion 62 that can travel on the rail 61 at the same speed in the billet B drawing direction. The cutting portion 62 includes a holding portion (not shown) that holds the billet B and a billet cutting cutter 63. And. The cutting unit 62 of the traveling cutting machine 6 starts cutting the billet B while traveling while holding the billet B in the vicinity of the target length. And if billet B is cut | disconnected, while holding | maintaining billet B, it will return to an initial position and it prepares for the cutting | disconnection of the next billet B. By using such a traveling cutting machine 6, the long material L can be obtained without stopping the casting process. On the other hand, the short cutter 7 is a device that cuts one long material L into a plurality of short cast bars S having a predetermined length, and a fixed type is used.

  8 is a processing machine for external cutting which performs the cutting process of the surface of the short cast rod S. The external cutting machine 8 externally cuts between 20 and 30 mm at both ends in the longitudinal direction of the short cast rod S. By performing this external cutting, it is easy to grip the short cast rod S at the time of peeling and to facilitate cutting with a cutting tool for peeling. This external cutting also has an effect of preventing forging defects such as cracks from occurring at both ends of the short cast rod S in a forging process performed later. If the irregularities (patterns) on the surface of the short cast rod S provided during the external cutting are determined as surface defects in the inspection process, the yield and efficiency of the process are greatly reduced as described above.

  9 is a peeling device for peeling the short cast rod S after the external cutting. The peeling device 9 includes a gripping part that grips both ends of the short cast bar S after cutting, and a cutting tool that turns the segregation layer on the surface of the short cast bar S.

  10 is a step of inspecting the surface and internal defects of the short cast rod S after peeling. In this step, the internal defect inspection of the short cast rod S is performed by a known ultrasonic flaw detection. In other words, the short cast rod S is irradiated with ultrasonic waves and the reflected waves are measured to check for flaws, internal defects, and the like. Although not shown, more specifically, it is composed of a water tank filled with running water, an ultrasonic sensor configured to be slidable with respect to the water tank, and a predetermined control device. The ultrasonic sensor irradiates ultrasonic waves while sliding along the longitudinal direction of the short casting rod S carried by the conveyor 81 one by one in a water tank filled with running water, so that the flaw detection inspection of the short casting rod S is performed. I do.

  On the other hand, in this inspection step 10, for example, after the ultrasonic flaw detection, a surface defect inspection of the short cast rod S is performed by a CCD camera. As for the order of ultrasonic inspection and CCD camera inspection in the inspection process, it is preferable to perform CCD camera inspection on the casting rod next to ultrasonic inspection.

  At this time, after wiping the dirt on the casting rod surface, a digital image of the casting rod surface was taken with a CCD camera, and this image was analyzed. The stain on the surface of the rod is distinguished from the actual surface defect of the cast rod, and is determined as a good product. The casting rod determined as non-defective is conveyed to the continuous heating furnace 12 by an automatic sorting device such as a robot 91. On the other hand, in this inspection process, what does not satisfy the predetermined standard is removed as scrap by an automatic sorting device and reused as a melting raw material.

  11 is a continuous heating furnace that heat-treats the short cast rod S that satisfies the predetermined standard after the inspection, and 12 is a forging device 12 that presses the short cast rod S after the heat treatment. . The forging device 12 is a plurality of press dies, for example, a bending die for bending a material, a buster die and a blocker die which are dies for rough forming forging, and a die for finishing to a final shape. It has a certain finisher mold.

  As described above, in the aluminum alloy forged part manufacturing apparatus 1, at least the processes from the continuous casting machine 5 to the automatic sorting apparatus in the inspection process 10 are automated. In such a large production line for automobile suspension parts, in order to assure the quality of the suspension parts that are safety parts, not only forged products but also for defects between the surface and the interior of the cast bar before forging are inspected and selected. It is inevitable.

  The manufacturing process of the forged part by such a manufacturing apparatus 1 is demonstrated below according to the flowchart of FIG. First, as step S1, a raw material such as an aluminum ingot or the scrap material and an alloy component element are introduced into the melting furnace 2, and these are melted to adjust the molten aluminum alloy so as to have a predetermined alloy component composition. To form.

  Next, component composition adjustment / removal refining / degassing processing is performed as step S2. In this process, in the melting furnace 2, a small amount of flux is added to the molten metal, and a refining process for removing the iron is performed. Further, after that, the molten metal is introduced into the degassing apparatus 4 through the holding furnace 3 to remove the hydrogen gas dissolved in the molten metal. Here, for the removal of the hydrogen gas, a known method such as a rotational degassing process using nitrogen gas or argon gas is employed. In the holding furnace 3, the temperature of the molten metal is also adjusted.

  In the subsequent continuous casting which is Step S3, the molten aluminum alloy is solidified in the continuous casting machine 5 and drawn out as billet B. The drawn billet B is guided by the casting conveyor 54 and sent to the traveling cutting machine 6 in the next process.

In step S <b> 4, the billet B is cut and cut into the long material L. The traveling cutting is performed by the traveling cutting machine 6 described above, and the long material L obtained by cutting is sent to the cross conveyor 71. The cross conveyor 71 conveys the billet B in the orthogonal direction, and the long material L carried out by the cross conveyor 71 is once stocked on the stock conveyor 72. Then, the long material L stocked on the stock conveyor 72 is discharged one by one toward the short cutting machine 7. In addition, generation | occurrence | production of an end material will be prevented when the length of the long material L cut | running | working by cutting in step S4 is made into the integral multiple of the length of the short cast rod S demonstrated later (a part for cutting is included). And the material yield can be improved.

  Further, in step S5, the long material L is short-cut by the short cutting machine 7 into a short casting rod S (for example, 20 to 70 cm in length) having a predetermined length according to the dimensional shape of the forged part. In the present embodiment, the aluminum alloy is cut into short cast rods S at this stage to improve the efficiency of the subsequent processing.

  The short cast rod S that has been cut into short pieces is transported by a transport conveyor 81, subjected to an external cutting process by the external cutting processing machine 8, and then peeled by a peeling device 9 as step S6.

  The short cast rod S after peeling is transported to the defect inspection step 10 by the transport conveyor 81. The short cast rod S transported to the defect inspection step 10 is washed away with industrial water such as fine powder chips and cutting oil adhering to the surface in the previous step. Thereafter, a defect inspection is performed (step S7), and the short cast rod S satisfying a predetermined standard is loaded on a storage container (not shown) by the robot 91 and conveyed to the next process.

  And the homogenization process of step S8 and the heating for forging of step S9 are performed by the continuous heating furnace 11. FIG. In addition, the homogenization process of step S8 and the heating for forging of step S9 correspond to a heat treatment process.

(Surface defect inspection by CCD camera)
Below, the aspect of the surface defect test | inspection by the CCD camera of the short casting stick | rod S in the test | inspection process 10 in FIG. 1 is demonstrated. FIG. 1 is an explanatory view showing a surface defect inspection mode of a short cast rod S by a CCD camera.

  In FIG. 1, S is a short cast rod to be inspected, 100 is a CCD camera, 101 is an image analysis means, and 102 is an image monitor. The casting rod S and the CCD camera 100 are relatively rotated and moved by a rotating and moving means (not shown), and the surface of the casting rod S is sequentially photographed over the entire circumference and the entire length of the casting rod S surface. .

  At this time, prior to the surface defect inspection by the CCD camera, it is essential to wipe off deposits on the surface of the casting rod, which are collectively referred to as dirt, such as oil, water droplets or other dirt on the surface of the casting rod S.

  As described above, after peeling, the short cast rod S transported to the defect inspection step 10 is washed with industrial water with adhering matter such as fine chips and cutting oil adhering to the surface in the previous step before ultrasonic flaw detection. To be removed. Even if a large amount of moisture adhering to the surface at this time remains as it is, it does not affect the ultrasonic flaw detection. For this reason, in the defect inspection process 10, first, the flaws adhering to the surface in the previous process and the adhering material such as cutting oil are washed with water, and then ultrasonic flaw detection for internal defect inspection is performed. Is preferred.

  In this case, if a lot of moisture adhering to the surface of the short casting rod S remains as it is, an error in detecting the surface defect of the casting rod is likely to occur for the surface defect inspection by the CCD camera. It is preferable to wipe off the surface and remove the remaining deposits such as moisture, fine chips, or cutting oil. Therefore, when the ultrasonic inspection and the CCD camera are used together as the inspection step 10 as in the present invention, it is preferable to perform the surface defect inspection by the CCD camera after the ultrasonic inspection for the internal defect inspection.

  As described above, even if the surface of the short cast rod S is wiped off before the surface defect inspection by the CCD camera, an error occurs in the surface defect inspection by the CCD camera due to moisture or other dirt that cannot be completely wiped off. This fact is also the subject of the present invention.

  FIG. 2 illustrates a means for wiping the surface of the short cast rod S before the surface defect inspection by the CCD camera. The wiping means has a configuration in which a nonwoven fabric 106 in contact with the surface of the short cast rod S, a porous sponge 105 made of a material such as cellulose or nylon, and a resin plate 104 are laminated and sandwiched by a sandwiching jig 107. As indicated by the arrows, the wiping means and the short casting rod S abut against each other while relatively moving and rotating, and moisture and other dirt are removed over the surface of the short casting rod S.

  The nonwoven fabric 106 and the sponge 105 have water absorbency, and the sponge 105 gives the nonwoven fabric 106 elasticity necessary for wiping. The resin plate 104 gives the nonwoven fabric 106 and the sponge 105 the rigidity necessary for wiping the surface of the short cast rod S. Thus, in order to automate this wiping operation, the wiping means is preferably configured to have water absorption, elasticity, and rigidity necessary for efficient wiping in a short time as described above.

(Image analysis)
The image of the surface of the casting rod S picked up by the CCD camera 100 is input to the image analysis means 101 and analyzed. In the following, an image analysis method (defect identification method) capable of discriminating minute irregularities on the surface of the casting rod, the various stains, and the like from actual surface defects (surface defects) of the casting rod will be described.

  In general, when the CCD camera 100 is a monochrome camera, the image analysis unit 100 pays attention to the luminance (brightness / darkness) signal level of the image signal, and when the CCD camera 3 is a color camera, the image analysis is performed. The means 100 pays attention to the luminance signal level and the color signal level of the image signal.

  However, in the defect identification by binarization processing that is usually used for image analysis by a CCD camera, the unevenness of the cast bar surface and the dirty part of the cast bar surface that are to be identified as good products in the present invention, Cannot be distinguished from actual surface defects.

(Binarization processing)
Usually, the actual defect portion on the surface of the object has lower brightness than the normal cast rod surface portion that is not. Therefore, if a certain density level is set and binarization processing is performed on the numerical data of the CCD camera 3, it is assumed that an image in which an actual defect portion on the subject surface is emphasized can be obtained. .

  In this binarization process, if the density (brightness difference) of each pixel is expressed in numerical order from 0 to 255, for example, in ascending order, a certain brightness and darkness range of 0 to 255 is obtained. A density value “0” for black having the lowest density is assigned to the value at or below the density value, and pure white having the highest density for values greater than this value. The density value “255” is assigned, and the density of all pixels is expressed by being divided into either two of 255 (pure white) or 0 (black). In other words, this is a method of an image processing method used when extracting only an image having a light / dark density (luminance difference) of a certain level or more or a certain level or less from a photographed image.

(Defect identification by color signal level)
Based on this binarization processing, the defect is identified by the color signal level of each pixel. For example, the highest luminance red is set to 0 and the lowest luminance black is set to 255, and the threshold is set between them. It is a method of providing and identifying. That is, in the case of a value smaller than or equal to this separation value, it is identified as irregularities on the surface of the cast bar accompanying the above-mentioned external machining and a dirt portion on the surface of the cast bar. Is a method for discriminating from the actual defective portion on the surface of the cast bar. However, in this method, there is no clear difference in the color signal level between the casting surface defect and the dirt (the color signal level is duplicated) regardless of the value of the above-mentioned separation. It tends to be unclear.

(Defect identification by the number of pixels)
Based on this binarization processing, the defect identification based on the number of pixels is performed by setting the light / dark density value “0” on the above binarization processing divided by a certain light / dark density value in the range of 0-255. Based on the number of arranged pixels, an actual defect portion on the surface of the casting rod is identified from the irregularities on the surface of the casting rod and the dirty portion on the surface of the casting rod due to the external cutting. For example, for this identification (discrimination), the number of array pixels to which the light / dark density value “0” is assigned is set to 20 or more. Then, when the number of arranged pixels in a group to which adjacent (adjacent) light and dark density values “0” are assigned is as large as 20 or more, it is identified as an actual defective portion on the surface of the cast bar. On the other hand, the unevenness of the cast bar surface and the dirty part of the cast bar surface due to the above-mentioned external cutting process have a case where the number of array pixels to which the density value “0” of light and darkness adjacent to each other (adjacent) is given is less than 20. To do.

  However, even with this identification method based on the number of pixels, if there is a difference in size between the casting defect and dirt, there is no clear difference between the size of the casting defect and dirt (overlapping). In other words, there is a problem that it cannot be identified as a surface defect. The actual defective part of the cast bar surface, the unevenness of the cast bar surface due to the external cutting and the dirty part of the cast bar surface are less bright and dark than the normal cast bar surface part, and the brightness (luminance signal) Level) is lowered.

  Therefore, as a premise of image analysis, unless the identification by the average value of the luminance difference between adjacent pixels, which will be described below, is performed, the unevenness of the cast bar surface and the dirty part of the cast bar surface due to the above-mentioned external machining, Cannot be distinguished from surface defects.

(Identification based on measured luminance average value of a group of pixels)
The present invention also uses the principle that the binarization process uses that the actual defect portion on the surface of the object has a lower luminance than the normal casting rod surface portion that is not. However, when using this principle, it is characteristic that the luminance average values of adjacent pixels are compared.

That is, in the image analysis by the CCD camera, the average luminance value of adjacent pixels is obtained from the measured luminance of each pixel among a group of adjacent pixels, and each of these obtained average luminance values is obtained. Comparison is made between each other, and those having a difference value of the compared luminance average values larger than the set value are identified as surface defects, while those having the decrease difference smaller than the set value are identified as non-defective products.

  The identification method for comparing the average luminance values of the adjacent pixels is based on the size of the irregularities on the cast bar surface and the dirt on the cast bar surface and the actual surface defects caused by the above-mentioned external cutting. First, the measured luminance of each pixel uses a point that there is a characteristic that a larger difference occurs in the average luminance value between adjacent pixels.

  That is, the brightness of the actual defect portion on the surface of the object, the unevenness on the cast rod surface accompanying the above-described external machining, and the dirty portion on the cast rod surface are similarly lower than the normal cast rod surface portion. However, in the actual surface defect, the difference in each measured luminance average value between the adjacent pixels is larger than the unevenness of the cast bar surface and the dirt on the cast bar surface due to the external cutting Become. In other words, the average brightness value of adjacent pixels in an actual surface defect changes steeply (abruptly) in comparison of the average brightness values of a plurality (group) of adjacent pixels. On the other hand, the unevenness of the cast bar surface and the dirt on the cast bar surface due to the above-described external machining have a difference in the clear tendency that the change is gentle.

Therefore, for each of the casting-forging conditions and casting-forging process conditions, the set value (threshold value) is determined by grasping the difference in the above tendency, and the difference value of the luminance average value compared is set value. Large ones are identified as surface defects, while if the difference value of the compared average brightness value is smaller than the set value, if the surface is stained or irregularities on the surface of the cast bar due to external machining, i.e. good products, Discrimination and selection can be performed easily and accurately.

This method of the present invention will be described more specifically with reference to the drawings. 3 (a) and 3 (b) show a group of (four) pixels among a large number of pixels lined up adjacent to each other at the time of image analysis by a CCD camera, respectively, and This is schematically illustrated. 3 (a) and 3 (b), the measurement brightness of each pixel, for example, the pixels that are arranged adjacent to each other linearly are denoted as A, B, C, and D in order from the left. In FIG. The luminance is 255 for A, 255 for B, 125 for C, and 0 for D. In FIG. 3B, the measured luminances are 255 for A, 255 for B, 0 for C, and 0 for D.

And the threshold value of the difference value of the luminance average value between the adjacent pixels is set to 125 or more, and the threshold value or more is set to be a surface defect portion, and the surface of the cast bar accompanying the external machining The uneven portion and the dirty portion of the cast rod surface are set to be less than this threshold value.

In the case of FIG. 3A, the value obtained by averaging the luminances of adjacent A and B is 255, the value obtained by averaging the luminances of adjacent B and C is 190, and the luminances of C and D are calculated. The average value is 62.5. The reduction rate (difference) between the average brightness value of A and B and the average brightness value of B and C is 65, and the average brightness value of B and C and the average brightness value of C and D The reduction rate is 127.5. Further, the difference between the average brightness value of A and B and the average brightness value of C and D is 192.5. Therefore, the threshold value is 125 or more between the average brightness value of B and C and the average brightness value of C and D, that is, between B and C. In this respect, A and B are high and the average value of both luminances is high and is a normal region. Therefore, the region between C and D having a difference value of the luminance average value larger than the threshold value. Are identified as surface defect portions.

  However, in reality, the region of C and D is not a surface defect but a stain. Therefore, even in this method, depending on the setting of the threshold value, the error is smaller than the binarization, but the detection accuracy may still be insufficient.

For this reason, in order to further improve the surface defect inspection accuracy, the difference value of the luminance average value larger than the threshold value 125 is continuously generated with two or more luminance average values in the group of pixels adjacent to each other. that identifies as surface defective. That is, it is not an identification based on whether or not the difference value of the average brightness value is greater than the threshold value 125, but whether the difference value of two or more brightness average values is continuously greater than the threshold value 125. that identifies with whether or not.

According to this, in the case of FIG. 3A, the threshold value is 125 or more only once, so that it is identified as a non-defective product. On the other hand, in FIG. 3B, each measured luminance is 255 for A, 255 for B, 0 for C, and 0 for D. A value obtained by averaging the luminances of adjacent A and B is 255, a value obtained by averaging the luminances of adjacent B and C is 127.5, and a value obtained by averaging the luminances of C and D is 0. Become. The decrease rate (difference) between the average brightness value of A and B and the average brightness value of B and C is 127.5, the average brightness value of B and C, and the average brightness value of C and D The reduction rate is 127.5. Therefore, the difference values of two or more luminance average values are continuously larger than the threshold value 125 for each luminance of A and B and the normal range where the average value of both luminances is also high. Are identified as surface defect portions.

  If the number of pixels of the pixel group for obtaining the average value of these luminance differences is, for example, the number of pixels in the range of 9 to 49 in a section of 3 to 7 in the vertical and horizontal directions, the identification can be performed more accurately. . However, if the number of pixels that average the brightness difference is too large or too small, the average value of the brightness difference will not reflect the actual surface defect part. Identification becomes difficult. For this reason, the average value of the brightness difference to be set and the number of adjacent pixels for obtaining the average value of the brightness difference itself are the actual defect portion of the actual cast bar surface and the cast bar surface due to the external cutting. It is preferable to determine and set the luminance difference and size, such as the uneven portion and the dirty portion of the cast rod surface.

  FIG. 4 shows a group of adjacent pixels at the time of image analysis of luminance measurement results by a CCD camera on the surface of an actual aluminum alloy cast bar manufactured in the steps of FIGS. In the figure, a group of pixels on the left side indicate actual surface defects (defective products), and a group of pixels on the right side are irregularities on the surface of the cast bar and dirt on the surface of the cast bar (non-defective product) due to the above-mentioned cutting.

  As shown in FIG. 4, when analyzed as the measured luminance of a group of adjacent pixels, the defective product on the left side and the non-defective product on the right side have a large reduction rate (in comparison with the average luminance value of adjacent pixels). There is a difference). That is, the change in luminance average value of the defective product on the left side is abrupt or steep at the periphery (periphery). On the other hand, the change in luminance of the right non-defective product is gentle from the center to the periphery (periphery).

  As described above with reference to FIG. 3, these points are quantified by comparing the difference between the luminance average values and the comparison with the threshold value. As described above, the left side of FIG. Both can be accurately identified as good products.

  Thus, in the forging line of an aluminum alloy casting rod, in order to distinguish the minute irregularities on the surface of the casting rod and the various kinds of dirt from the surface defects (surface defects) of the actual casting rod, a CCD camera is used. It can be seen that discrimination is necessary based on the binarization processing for the numerical data of the image, the luminance difference between adjacent pixels based on this, and the number of adjacent array pixels whose light and dark densities are above a certain level. Therefore, even if a high-resolution visible light CCD camera with a resolution smaller than 1000 μm as in Patent Document 3 described above is used, such binarization processing, the number of adjacent array pixels, and the luminance between adjacent pixels If there is no discrimination based on the average value of the differences, it can be seen that it is difficult to discriminate between the two.

  7 and 8, the surface defect inspection of the short aluminum alloy cast round bar S in the inspection step 10 using a CCD camera and an image analysis system (XV-1000 series manufactured by Keyence Corporation) is performed. The results obtained are described below.

  FIG. 5 compares the luminance average values of adjacent pixels in the range of 4 × 4 pixels in the adjacent pixels in FIG. 4, and the luminance average value difference is 100. This is an example in which the set value is automatically determined as 100 or more defective products and less than 100 as non-defective products. Note that this set value is determined for each round bar manufacturing process by collecting data on a number of luminance average value differences between defective products and non-defective products.

  In the pixel on the left side of FIG. 5, two or more luminance average value differences (decrease rates) are consecutively 127.5, which is larger than the threshold value 125, and are identified as defective products. However, FIG. 5 itself does not indicate that two or more luminance average value differences (decrease rates) larger than the threshold value are consecutive. On the other hand, the luminance average value difference (decrease rate) of the right pixel is 30.1 which is continuously smaller than the threshold value 125, and is identified as a non-defective product. According to the comparison of the luminance average values of the adjacent pixel groups as described above, since a large luminance average value difference is actually produced between the defective product and the non-defective product, both can be identified.

  On the other hand, FIG. 6 shows the result of binarizing the same pixel group in FIG. 4 and identifying the defect based on the number of pixels, and binarizing with a threshold value of 127.5. In this example, an area (number of pixels) of 3000 or more is defined as a defect (defective product). The number of pixels on the left side is 3397, the number of pixels on the right side is 4182, and the number of pixels of the non-defective product and the defective product is close to each other, so that both are identified as defective products. The same applies to the identification at the color signal level described above.

  Therefore, from these results, unless the discrimination is made by comparing the luminance average values of adjacent pixels in a group of pixels according to the present invention, the irregularities on the surface of the cast bar due to the above-mentioned external machining are classified into the category of dirt on the cast bar surface. It can be recognized (identified) that it cannot be distinguished from the actual surface defect of the cast bar and cannot be distinguished as a non-defective product.

  As described above, according to the present invention, an aluminum alloy bar material that can distinguish minute irregularities on the surface of the aluminum alloy bar material, various kinds of dirt, and the like from surface defects (surface defects) of the actual bar material. A defect inspection method can be provided. As a result, it is suitable for a casting line of an aluminum alloy casting rod, a casting-forging line, an aluminum alloy forging line, or the like, and it is possible to perform defect inspection without increasing or reducing the yield and production efficiency of these production lines.

It is explanatory drawing which shows the aspect of the surface defect inspection by the CCD camera of a casting rod. It is explanatory drawing which shows the one aspect | mode of the wiping means of the dirt on a casting stick surface. It is explanatory drawing which shows the technical idea of this invention. It is explanatory drawing which shows the pixel group of the actual stick | rod surface by the CCD camera of an Example. It is explanatory drawing which shows the identification result by the example of invention of FIG. It is explanatory drawing which shows the identification result by the comparative example of FIG. It is a whole lineblock diagram of a manufacturing line of aluminum alloy forging parts. It is a flowchart which shows the manufacturing process of an aluminum alloy forge part. It is explanatory drawing which shows the actual surface defect of a casting rod, and dirt on the casting rod surface.

Explanation of symbols

S: casting rod, 100: CCD camera, 101: image analysis means,
102: Image monitor, 103: Wiping means, 104: Resin plate,
105: Porous sponge, 106: Non-woven fabric, 107: Holding jig

Claims (3)

In the method for inspecting the surface defect of the manufactured aluminum alloy bar, after cleaning the surface of the bar, the brightness of the bar surface is measured with a CCD camera, and the image analysis with this CCD camera is performed. among the group of adjacent pixels, the two pixels adjacent to each other from the measured luminance of each pixel After each calculated brightness average value, further, the two luminance average value of adjacent pixels, the two adjacent The difference between the luminance average values of pixels adjacent to one of the pixels in the opposite direction (hereinafter referred to as “difference value of luminance average value”) is obtained, and the difference value of the luminance average value is larger than the set value. while identifying which occurs two or more times in succession to the surface defective in a group of adjacent pixels said is either the difference value of the average brightness value is less than all the set value, or the average luminance value Surface defect inspection method of an aluminum alloy bars, characterized in that that the difference value is greater than the set value identified as good product which does not occur in succession in a group of adjacent pixels said. The surface defect inspection method for an aluminum alloy bar according to claim 1, wherein the aluminum alloy bar is an aluminum alloy cast bar . An aluminum alloy cast bar is cast as an aluminum alloy bar, and after machining the cast bar, defects between the surface and the inside of the cast bar are inspected and selected and hot forged to produce an aluminum alloy forged material. Aluminum in the step of automatically performing the steps from casting of the aluminum alloy cast rod to defect inspection and selection of the cast rod, and performing internal defect inspection of the cast rod by ultrasonic flaw detection. In the method of inspecting the surface defect of the alloy cast bar, the brightness of the cast bar surface is measured with a CCD camera after the dirt on the surface of the aluminum alloy cast bar is wiped off. After obtaining the average luminance value of two adjacent pixels from the measured luminance of each pixel, the pixels The difference between the average luminance value of the two pixels and the average luminance value of the adjacent pixels in the opposite direction to one of the two adjacent pixels is obtained, and the difference between the average luminance values is obtained. A value that is greater than a set value is identified as a surface defect of the casting rod that occurs two or more times consecutively in the adjacent group of pixels, while the difference value of the average brightness value is all smaller than the set value. Or the unevenness of the surface of the cast bar or the cast bar that accompanies the external cutting process that does not continuously occur in the adjacent group of pixels that the difference value of the luminance average value is larger than the set value. A method for inspecting a surface defect of an aluminum alloy bar, wherein the surface defect is identified as dirt on the surface.

Images