How to detect product defects

Abstract

To provide a defect detection method which can detect a blow hole of a casting easily.SOLUTION: A defect detection method comprises the steps of: acquiring a cross section image obtained by cutting in round slices a prescribed number of same-shaped castings at the same position; calculating the average luminance obtained by averaging the luminance of each corresponding image area of the cross section images and generating a reference cross section image made of the respective image areas applied with the average luminance; acquiring a detection cross section image obtained by cutting in round slices a casting having the same shape as the casting becoming the defect detection object at the same position; calculating the difference luminance taking a difference in the luminance between each image area of the detection cross section image and each image area of the reference cross section image equivalent to each image area and generating a cross section image for defect detection made of the image areas applied with the difference luminance; and specifying a blow hole from the cross section image for defect detection.SELECTED DRAWING: Figure 8

Description

[Technical field]

[0001] The present invention relates to a method for detecting defects in a product, and more particularly to a method suitable for detecting blowholes, which are internal defects that occur in castings. [Background technology]

[0002] Attempts have been made to use CT (computed tomography), which can obtain cross-sectional images, to detect defects in products, such as blowholes that occur in castings (Patent Document 1), but CT images have a particular type of noise called artifacts, which causes a problem of reduced defect detection accuracy.For example, Patent Document 2 shows a method for removing metal artifacts that occur in medical CT images, in which a circular detection area is set with a pixel point of a metal area as its center, and when the ratio of the metal area to the non-metal area in the circular detection area is equal to or greater than a preset value, the circular detection area is used as a boundary area and boundary smoothing processing is performed to remove the metal artifacts. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2005-351875 [Patent Document 2] Patent Publication 2015-85198 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above-mentioned conventional artifact removal methods require complex calculations, and therefore have the problem that product defects cannot be easily detected.

[0005] Therefore, the present invention is intended to solve such problems, and aims to provide a defect detection method that can reduce the computational burden of removing artifacts from CT images and easily detect product defects. [Means for solving the problem]

[0006] In order to achieve the above object, the first invention obtains cross-sectional images of a predetermined number of products of the same shape cut at the same position, calculates an average brightness by averaging the brightness of each corresponding image area of ​​these cross-sectional images, and generates a reference cross-sectional image consisting of each of the image areas to which this average brightness has been added; on the other hand, obtains detection cross-sectional images of a product of the same shape as the product to be subjected to defect detection cut at the same position, calculates a differential brightness by taking the difference between the brightness of each image area of ​​the detection cross-sectional image and each image area of ​​the reference cross-sectional image corresponding to each image area, and generates a cross-sectional image for defect detection consisting of the image areas to which this differential brightness has been added. A required number of cross-sectional images for defect detection are generated at different cross-sectional positions of the product, and a three-dimensional image for defect detection is generated from the cross-sectional images for defect detection. The method is characterized in that defects are identified from the image. Note that the "image area" may be an area of ​​one pixel, or an area of ​​a predetermined number of pixels.

[0007] In the first invention, the reference cross-sectional image is assigned an average brightness obtained by averaging the brightness of the corresponding image areas of the cross-sectional images obtained by cutting a predetermined number of identical products at the same position. Therefore, even if there is a local defect inside each of the products and the brightness changes significantly at this part, the brightness change due to the defect is diluted and becomes sufficiently small by averaging, and a reference cross-sectional image without a defective image part is obtained. In this case, since the artifacts on each of the cross-sectional images always appear at the same position and with the same brightness if the installation position and orientation of the identical products are the same, they remain on the reference cross-sectional image even if the brightness is averaged as described above. On the other hand, the detection cross-sectional image appears with the defective image part and the same artifacts superimposed thereon. Therefore, by calculating the difference in brightness between each image area of ​​the detection cross-sectional image and each image area of ​​the reference cross-sectional image corresponding to each image area, a defect detection cross-sectional image consisting of image areas to which the difference brightness is assigned is generated, and the artifacts are removed from the defect detection cross-sectional image and only the image of the defect remains, so that the defect can be easily and reliably identified. In this way, according to the first invention, product defects can be easily detected by simple calculations, such as averaging luminance values ​​and calculating luminance differences, without requiring complex calculations. In addition, since a three-dimensional image for defect detection is generated, the size (volume) and location of the defect can be detected more accurately, and the defect can be identified more reliably.

[0010] The second invention Now, obvious defects are removed in advance from the predetermined number of cross-sectional images acquired.

[0011] The second invention According to the method, obvious defects are removed beforehand from each cross-sectional image by automatic or visual inspection, etc., and the efficiency of identifying remaining defects is improved.

[0012] The third invention In the above, the product is a casting and the defect is a blowhole. Effect of the Invention

[0013] As described above, according to the defect detection method of the present invention, the computational burden of removing artifacts from a CT image is reduced, and product defects can be easily detected. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a schematic diagram of an apparatus for carrying out the method of the present invention. [Diagram 2] 10 is a flowchart showing a procedure for generating a reference cross section image. [Diagram 3] 1A to 1C are diagrams showing examples of cross-sectional images of a casting at three positions in the Z direction. [Figure 4] 1A to 1C are diagrams showing examples of cross-sectional images at three positions in the Z direction of a CAD three-dimensional image of a casting. [Diagram 5] 10 is a flowchart showing a detailed procedure of position correction. [Figure 6] FIG. 13 is a diagram showing a corner of a CAD cross-sectional image and a corner center of gravity of a CT cross-sectional image. [Figure 7] 10 is a flowchart showing a detailed procedure for calculating an average luminance. [Figure 8] 10 is a flowchart showing a procedure for generating a cross-sectional image for defect detection. [Figure 9] 10 is a flowchart showing a detailed procedure for calculating a brightness difference. [Figure 10] 10 is a flowchart showing a detailed procedure for removing edge noise. [Figure 11] 10 is a flowchart showing a detailed procedure for identifying blowholes. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] It should be noted that the embodiment described below is merely an example, and various design improvements made by those skilled in the art without departing from the gist of the present invention are also included in the scope of the present invention.

[0016] An example of an apparatus for carrying out the method of the present invention is shown in Fig. 1. For example, a casting W as a product is placed on a table (not shown) that rotates in the XY plane. An X-ray irradiator 1 and a line sensor 2 that constitute a CT (computed tomography) are positioned on opposite sides of the casting W in a plane parallel to the XY plane, and X-rays R that spread out in a fan shape, with the beam narrowed to a flat shape by a collimator or the like, are emitted from the X-ray irradiator 1 and penetrate the casting W in a cross-sectional manner, and are input to the line sensor 2.

[0017] The output from the line sensor 2 is input as digital data to a computer 3 in the control device. By rotating the casting W on the table, transmitted X-rays from all directions are input to the line sensor 2, and the computer 3 inputs the digital data from the line sensor 2 and obtains a cross-sectional image of the casting W at a predetermined position in the Z direction by a known method.

[0018] In this embodiment, the table on which the casting W is placed, the X-ray irradiator 1, and the line sensor 2 are relatively movable in the Z direction, and a three-dimensional image of the casting W can be obtained from cross-sectional images obtained at multiple positions in the Z direction by calculations in the computer 3. Rotation of the X-ray irradiator 1 and the table, relative movement in the Z direction, etc. are appropriately performed by output signals from the computer 3. Note that each procedure described below in the flowchart is executed by a program of the computer 3.

[0019] A. Generation of reference cross-sectional images (Acquisition of cross-sectional images) FIG. 2 shows a procedure for generating a reference cross-sectional image. First, in step 101, a cross-sectional image of a casting at a predetermined Z-direction position is acquired. In this embodiment, cross-sectional images of one casting at multiple positions in the Z direction are acquired. This is shown in FIG. 3, where F1, F2, and F3 are examples of cross-sectional images of a casting W acquired at three positions Z1, Z2, and Z3 in the Z direction, respectively. In this embodiment, after cross-sectional images of one casting at multiple positions in the Z direction (a predetermined number required to obtain a three-dimensional image described later) are acquired, cross-sectional images are acquired at the same positions in the Z direction for a required number of other castings of the same shape.

[0020] (Correction of cross-sectional image position) Next, in step 102 in Fig. 2, the positions of the cross-sectional images F1 to F3 are corrected. This is because the position of the casting placed on the table shifts slightly each time. In this embodiment, this shift does not occur in the Z direction because the casting is placed on the table, but occurs only in the XY plane. Note that if the position shift when the casting is placed is small enough to not cause any problems, there is no need to correct the position.

[0021] In this embodiment, the position correction of the cross-sectional image is performed by using a separately prepared CAD three-dimensional image of the casting. Fig. 4 shows an example of a CAD three-dimensional image W', in which F1', F2', and F3' are CAD cross-sectional images obtained by slicing the CAD three-dimensional image W' at Z-direction positions Z1', Z2', and Z3', which are the same as the above-mentioned three positions Z1, Z2, and Z3.

[0022] Details of the position correction step 102 are shown in Fig. 5. In Fig. 5, in step 201, the CT cross-sectional image data (corresponding to the cross-sectional image) and the corresponding CAD cross-sectional image data (corresponding to the CAD cross-sectional image) are extracted. In step 202, the CAD cross-sectional image data is scanned to detect the four corners. The corners C1', C2', C3', and C4' detected in the CAD cross-sectional image F1' are shown in Fig. 6(1).

[0023] In step 203 in Fig. 5, the vertical and horizontal differential values ​​of the CT cross-sectional image data corresponding to the periphery of the detected corners C1' to C4' are calculated, and the centers of gravity having differential values ​​equal to or greater than a certain amount are calculated (step 204). Fig. 6(2) shows the calculated centers of gravity C1, C2, C3, and C4 on the cross-sectional image F1.

[0024] 5, the amount of translation and rotation of the CT cross-sectional image data relative to the CAD cross-sectional image data is calculated based on the relative positional relationship between the corners C1'-C4' and the centers of gravity C1-C4. This calculation is performed, for example, by determining an appropriate amount of translation and rotation angle that minimizes the least square error using the relative positional relationship between the corners C1'-C4' and the centers of gravity C1-C4 detected or calculated at multiple positions Z1-Z3 in the Z direction of the casting. In step 206, the CT cross-sectional image data is moved by the calculated amount of translation and rotation angle to correct the image positions of the cross-sectional images F1-F3.

[0025] (Calculation of average brightness, generation of reference cross-sectional images) In step 103 in Fig. 2, the average brightness of each image region (pixels in this embodiment) is calculated for the position-corrected cross-sectional images F1 to F3. This calculation of the average brightness is carried out by calculating the average brightness of each image region of each cross-sectional image sliced ​​at the same position in the Z direction for a required number of castings W of the same shape. By calculating the average in this manner, even if there is a cast cavity as a local defect inside the casting and the brightness changes significantly in this area, the averaging dilutes the brightness change due to the cast cavity and makes it sufficiently small.

[0026] On the other hand, since artifacts always appear at the same position and with the same brightness in each cross-sectional image, they remain as they are even if the average brightness is calculated as described above. Therefore, when the brightness of each image region is averaged and a reference cross-sectional image is generated from the image regions to which such average brightness has been assigned (step 104), the reference cross-sectional image is one in which artifacts are superimposed on a cross-sectional image of the casting from which the blowholes have been eliminated. In this embodiment, such reference cross-sectional images are generated at multiple positions in the Z direction of the casting (the number of positions required to obtain a three-dimensional image, which will be described later).

[0027] Details of the average brightness calculation step 103 (FIG. 2) are shown in FIG 7. In step 301 in FIG 7, the CT value (corresponding to the brightness value) of the reference cross-sectional image data is reset to 0. Next, the CT value of the CT image data is added to the reference cross-sectional image data (step 302), and this addition is repeated for the required number of CT cross-sectional image data (step 303), and in step 304, the CT values ​​of all pixels of the reference cross-sectional image data are divided by the number of CT cross-sectional image data (required number).

[0028] The calculation of the average luminance is not necessarily limited to the arithmetic mean as described above, but may be a weighted average or the like as necessary.

[0029] B. Generation of cross-sectional images for defect detection (Acquisition of detected cross-sectional image) Figure 8 shows the procedure for generating a cross-sectional image for defect detection. First, in step 401, a detection cross-sectional image is acquired. The detection cross-sectional image is a cross-sectional image cut in a predetermined position in the Z direction obtained by placing a casting that is to be subjected to defect detection on a table and performing the above-mentioned CT. The casting that is to be subjected to defect detection has the same shape as the casting for which the reference cross-sectional image has been generated.

[0030] In step 402 in FIG. 8, the position of the detected cross-sectional image is corrected in the same manner as described above (see step 102 and steps 201 to 206).

[0031] (Calculation of differential brightness) Next, in step 403 of FIG. 8, the difference in brightness between the detected cross-sectional image and the reference cross-sectional image is calculated. The reference cross-sectional image in this case is obtained by cutting the casting in a cross-sectional view at the same position in the Z direction as the detected cross-sectional image. Here, the detected cross-sectional image is a cross-sectional image of the casting on which artifacts and images of blowholes (if any) are superimposed. In contrast, the reference cross-sectional image is, as described above, a cross-sectional image of the casting on which artifacts are superimposed. Therefore, by calculating the difference in brightness between the detected cross-sectional image and the reference cross-sectional image, a defect detection cross-sectional image can be obtained in which the artifacts and the cross-sectional image portion of the casting are removed and only the image portion of the blowhole remains.

[0032] Details of the differential brightness calculation step 403 are shown in Fig. 9. In step 501 in Fig. 9, the CT value of the reference cross-sectional image data is subtracted from the CT value of the CT cross-sectional image data (corresponding to the detected cross-sectional image), and if the subtraction result is a positive value, it is set to 0 (step 502), and if the subtraction result is a negative value, it is multiplied by -1 to make it a positive value (step 503). If there is a cast cavity, the brightness of that part will decrease, so the subtraction value will be a large negative value. Therefore, this is converted to a positive value.

[0033] (Removal of low brightness difference areas) In step 404 in FIG. 8, any portion of the defect detection cross-sectional image obtained in step 403 that has a brightness lower than a threshold value is removed as noise.

[0034] (Removal of edge noise, generation of cross-sectional images for defect detection) In step 405 in Fig. 8, high luminance parts occurring along the contour of the casting are removed as edge noise in the defect detection cross-sectional image to generate a defect detection cross-sectional image (step 406). Fig. 10 shows the details of the edge noise removal step 405. In Fig. 10, a boundary distance is set in step 601, and in step 602, if there is non-zero difference data (difference luminance value) within an area less than the boundary distance from the contour of the casting that can be recognized in the CAD cross-sectional image data corresponding to the defect detection cross-sectional image data, this is set to 0.

[0035] (Generating 3D images for defect detection, identifying blowholes) In step 407 of Fig. 8, a required number of cross-sectional images for defect detection are obtained at different cross-sectional positions in the Z direction of the casting to generate a three-dimensional image for defect detection. This three-dimensional image for defect detection is one in which only the blowhole parts (candidates) are arranged in three-dimensional space. In the following step 408, blowholes are identified from the blowhole candidates in the three-dimensional image for defect detection.

[0036] Details of the blowhole identification step 408 are shown in Fig. 11. In step 701 in Fig. 11, blowhole determination criteria (volume, length, location) are set. In step 702, blowhole candidates that are continuous non-zero regions are detected in the three-dimensional image data for defect detection. In the following step 703, the volume, length, and location of the detected blowhole candidates are calculated, and if the blowhole determination criteria are met, the blowhole candidate is identified as a blowhole (steps 704 to 705). This procedure is performed for all blowhole candidates (step 706).

[0037] (Other embodiments) In the above embodiment, a blowhole is identified using a three-dimensional image for defect detection. However, a blowhole may be identified using a two-dimensional cross-sectional image for defect detection. In the above embodiment, if obvious defects in each cross-sectional image are removed in advance by visual inspection or the like, the remaining defects can be efficiently identified. Although the above embodiment is directed to detecting blowholes in a casting, the present invention can be widely applied to detecting defects in products. [Explanation of symbols]

[0038] 1...X-ray irradiator, 2...line sensor, 3...computer, R...X-ray, W...casting product (product).

Claims

1. a detection cross-sectional image of a product having the same shape as the product to be subjected to defect detection, the detection cross-sectional image being obtained by cutting a predetermined number of products of the same shape at the same position, calculating an average brightness by averaging the brightness of each corresponding image area of ​​these cross-sectional images, and generating a reference cross-sectional image consisting of each of the image areas to which this average brightness has been added; a detection cross-sectional image of a product having the same shape as the product to be subjected to defect detection, the detection cross-sectional image being obtained by cutting a product having the same shape as the product to be subjected to defect detection, the detection cross-sectional image being obtained by cutting the product at the same position, calculating a differential brightness by taking the difference between the brightness of each image area of ​​the detection cross-sectional image and each image area of ​​the reference cross-sectional image corresponding to each image area, and generating a defect detection cross-sectional image consisting of the image areas to which this differential brightness has been added;

2. 2. The method for detecting defects in products according to claim 1, further comprising the step of removing obvious defects in advance from each of the predetermined number of acquired cross-sectional images.

3. 3. The method for detecting defects in a product according to claim 1, wherein the product is a casting, and the defect is a blowhole.

Images