JP2019049507A - Method and apparatus for detecting defects - Google Patents

Abstract

An object of the present invention is to provide a defect detection method capable of reliably detecting a defect of a complex-shaped inspection object without requiring much labor by masking.
An outline of an inspection object 1 is detected, and an outline portion in which a difference value between an outline approximation line L1 closest to the outline and each outline point on the outline becomes a predetermined value Th or more is detected as a defect 11. Do. The defect 11 is a convex defect generated on the blade edge of the turbine blade.
[Selected figure] Figure 2

Description

  The present invention relates to a defect detection method and a defect detection apparatus suitable for detecting a defect such as a convex defect or a dent of an inspection object.

  For example, in the case of a complex-shaped inspection object such as a turbine blade of a precision casting product, the complex shading causes a large difference in the luminance of each part, so repeated removal of non-uniform areas of luminance by masking is performed. There is a problem that it is necessary to create a section, which requires a lot of time and that an uninspected area is generated by masking. In addition, about the production | generation of the division of the defect detection area | region by masking, it describes in patent document 1, for example.

JP 2011-65604 A

  The present invention solves such a problem, and provides a defect detection method and a defect detection apparatus capable of reliably detecting a defect on an inspection object having a complicated shape without requiring much labor by masking. The purpose is to

  In the defect detection method of the present invention, the contour of the inspection object (1) is detected, and the difference value between the contour approximation line (L1) closest to the contour and each contour point on the contour is a predetermined value (Th) or more The contour portion that has become is detected as a defect (11).

According to the defect detection method of the present invention, a portion where the difference value between each contour point on the contour of the inspection object and the contour approximation line is equal to or more than a predetermined value is detected as a defect. Defects can be detected reliably without the need for complicated work such as area removal.
The contour approximation closest to the contour may be selected from a linear approximation or a curve approximation.
It is possible to set an inspection range for each of the straight line portion and the curve portion among the detected contours, and to select a linear approximation line or a curve approximation line in these inspection ranges.
When the difference value exceeds a predetermined threshold Th, the contour portion can be detected as a defect.
The threshold is, for example, 1.5% to 3.5%.

  In the defect detection apparatus of the present invention, a means (step 101, 102) for detecting the contour of the inspection object (1), and a means (step 103) for setting a contour approximation line (L1) closest to the contour; A means (step 104) for calculating a difference value between the contour approximation line and each of the contour points on the contour, and a means for detecting as a defect (11) a contour portion where the difference value is equal to or greater than a predetermined value (Th) Step 105, 106). The same effects as the above-described defect detection method can be obtained by the defect detection device of the present invention.

  The contour approximation line (L1) may be obtained from the contour of the non-defective inspection object (1). Further, the defect (11) is, for example, a convex defect or a concave defect generated on a blade edge of a turbine blade.

  The reference numerals in the parentheses indicate the correspondence with the specific means described in the embodiments to be described later.

  As described above, according to the present invention, it is possible to reliably detect a defect on an inspection object having a complicated shape without requiring much labor by masking.

It is a partial perspective view of a turbine blade which produced a concave defect in a 1st embodiment. It is a partial perspective view of a turbine blade which produced a concave defect. It is a figure which shows an example of an outline and an outline approximation line. It is a partial perspective view of a turbine blade which produced a convex defect. It is a partial perspective view of a turbine blade which produced a convex defect. It is a flowchart which shows the processing procedure in a computer. It is a flowchart which shows the processing procedure in a computer in 2nd Embodiment. It is a figure which shows a part of FIG.

  The embodiments described below are merely examples, and various design improvements made by those skilled in the art without departing from the scope of the present invention are also included in the scope of the present invention. Note that each step of the flowcharts of FIG. 6 and FIG. 7 described below is executed in a computer that has captured an image.

First Embodiment
A part of the image of the turbine blade 1 which is a test object is shown as an example in FIG. In the present embodiment, first, the turbine blade 1 is imaged by a known imaging unit (step 101 in FIG. 6), and in the image, an inspection range X1 set for a linear outline (set of outline points) Contour approximation lines L1 and L2 are set on the blade edge of the turbine blade 1 for each inspection range X2 set for a curved contour (step 103 in FIG. 6, FIG. 1 (b)). The contour approximation lines L1 and L2 are obtained by selecting an approximation line closest to the contour of the blade edge of the turbine blade 1 in each inspection range X1 and X2 from a linear approximation line or a curve approximation line. Here, as an example, a straight line approximation line is selected as the contour approximation line L1, and a curve approximation line is selected as the contour approximation line L2. The inspection ranges X1 and X2 are respectively divided into a linear portion and a curved portion of the contour of the turbine blade 1 and set.

On the other hand, the outline of the turbine blade 1 which is the inspection object is calculated from the captured image (step 102 in FIG. 6), and each outline of the blade edge and the inspection area for each inspection area X1 and X2 in the image The difference value (%) of the contour approximation lines L1 and L2 preset in X1 and X2 is calculated (step 104 in FIG. 6). When the difference value exceeds a predetermined threshold Th, it is determined that there is a defect (steps 105 and 106 in FIG. 6). If the difference value does not exceed the threshold Th, no defect is made (steps 105 and 107 in FIG. 6). Here, as shown in FIG. 3, the difference value is expressed by the following equation (1) or (2) from the distance h from the reference line PL to the contour approximation line L and the distance h ′ from the reference line PL to the contour line P: Calculated using selective use.
In the case of h '> h, difference value (%) = (h'-h) / h × 100 (1)
In the case of h> h 'difference value (%) = (h−h ′) / h × 100 (2)

  Here, the reference line PL is determined as follows. That is, in the captured image, the horizontal direction is the X axis (position), and the vertical direction is the Y axis (distance) (FIG. 8C), and the overall size (distance) of the inspection object on the image in the Y axis direction is The position of the half W / 2 when it is W is used as a reference (FIG. 8 (a)). Then, for the plot point S on the contour approximate line L1, a point near the plot point S at a distance of, for example, 90% of the distance W / 2 is selected as the reference point P. In this manner, in the inspection range X1, a plurality of reference points (P1, P2, P3) are obtained in the X axis (position) direction, and a combination of these is set as a reference line PL (FIG. 8 (b), (C)). The reference point P can be appropriately determined between 80% and 90% of the distance W / 2. Further, the inspection ranges X1 and X2 (FIG. 1) are merely examples, and in practice, many more are set at necessary places.

  Here, for example, as shown in the A region of FIG. 1 and FIG. 2, when the concave defect 11 is generated at the blade edge of the inspection range X1, the equation (2) is selected. In the concave defect portion 11, the difference value between the contour point and the contour approximate line L1 locally becomes larger than the threshold value Th. For example, when the threshold value Th is 2.5%, the difference value becomes 5.0% in the defect 11 portion of the A region in FIG. 1, and it is detected that the defect 11 is present in this portion (step 106 in FIG. 6). ). In the present invention, the threshold Th is set to 2.5% based on the experimental result that defects such as convex defects and dents of the inspection object in the present invention are significantly detected with a difference value of around 2.5%. . A suitable threshold Th is 1.5% to 3.5%.

  As another example, as shown in area B of FIG. 4 and FIG. 5, in the inspection ranges X1 and X2 in the image of the turbine blade 1 to be inspected, a contour approximation line is drawn to the blade edge on the opposite side in the thickness direction. If L3 and L4 are set, if a convex defect 12 due to a dent is generated at the blade edge in the inspection range X2, the difference between the contour point and the contour approximate line L4 at this convex defect 12 portion is a predetermined value Th Since it locally grows beyond it, it is detected that there is a defect 12 in this part. For example, when the threshold value Th is 2.5%, the difference value becomes 5.5% at the defect 12 portion in the A region of FIG. 4 and it is detected that the defect 12 is present in this portion

Second Embodiment
In the first embodiment, the turbine blade to be inspected is imaged and the contour approximation line is set from this, but in the present embodiment, as shown in steps 201 to 203 of FIG. From this, the outline is calculated and the outline approximation is set. Steps 204-209 are the same as steps 101, 102, 104-107 of FIG. 6 in the first embodiment. According to such a procedure, the defect detection accuracy can be further improved.

1 Turbine blade (object to be inspected) 11, 12 Defect, L1, L2, L3, L4 Contour approximate line.

Claims (8)

A defect detection method for detecting an outline of an inspection object and detecting an outline portion in which a difference value between an outline approximation closest to the outline and each outline point on the outline is equal to or more than a predetermined value as a defect. The defect detection method according to claim 1, wherein the contour approximation line is obtained from the contour of the inspection object without defects. The defect detection method according to claim 1, wherein the defect is a convex defect or a concave defect generated on a blade edge of a turbine blade as the inspection object. The defect detection method according to any one of claims 1 to 3, wherein a contour approximation line closest to the contour is selected from a linear approximation line or a curve approximation line. The defect detection method according to any one of claims 1 to 4, wherein an inspection range is set for each of a linear portion and a curve portion among the detected contours, and a linear approximation line or a curve approximation line is selected in each of these inspection ranges. Method. The defect detection method according to any one of claims 1 to 5, wherein the contour portion is detected as a defect when the difference value exceeds a predetermined threshold value Th. The defect detection method according to claim 6, wherein the threshold is 1.5% to 3.5%. A means for detecting the contour of the inspection object, a means for generating a contour approximation line closest to the contour, a means for calculating a difference value between the contour approximation line and each contour point on the contour, and the difference value And a means for detecting an edge portion having a predetermined value or more as a defect.