JP5729235B2 - Weld bead defect detection apparatus and weld bead defect detection method - Google Patents

Description

  The present invention relates to a technique of a weld bead defect detection device and a weld bead defect detection method for detecting a defect of a weld bead.

  A weld bead defect detection method is known in which a cross-sectional shape of a weld bead is measured as a shape point cloud by a light cutting method, and a region where the point cloud is missing in the measured shape point cloud is detected as a defect site such as a hole. (For example, patent document 1).

  However, in a defective part such as a hole, an abnormal value of the shape point group may occur due to multiple reflection of laser light during measurement. Further, the abnormal value of the shape point group generated in the perforated part may be a shape point group that covers the perforated part. The shape point group that covers the perforated part is not detected as the perforated part, and therefore, the determination is performed by overlooking the perforated part.

  On the other hand, an abnormal value of the shape point group due to multiple reflection of laser light may occur even in a normal part without any defect, for example, a corner between the base material and the bead or a plate-matching part with high reflectivity. . Detecting an abnormal value at a normal part without any defect as a defective part determines that a non-defective product is abnormal.

  Therefore, in the weld bead defect detection device and the weld bead defect detection method, it is necessary to prevent the perforation from being missed and to minimize erroneous detection. That is, in the weld bead defect detection device and the weld bead defect detection method, it is necessary to improve the detection accuracy of the weld bead defect.

JP 2008-215839 A

  The problem to be solved by the present invention is to provide a weld bead defect detection device and a weld bead defect detection method capable of improving the detection accuracy of weld bead defects.

  The problem to be solved by the present invention is as described above. Next, means for solving the problem will be described.

That is, in claim 1, a weld bead defect detecting device for detecting a defect of a weld bead, the point cloud data acquiring means for acquiring point cloud data in a predetermined cross-sectional direction for the surface shape of the weld bead, and the point cloud Based on a comparison between the data and a smooth model obtained by smoothing the point cloud data, or a comparison between the smooth model and a broken line model in which the surface shape of the weld bead is approximated to a broken line using the smooth model , Based on point cloud data classification means for classifying point cloud data into abnormal points and normal points, the presence or absence of the abnormal points between the adjacent normal points, and the distance between the adjacent normal points, The point cloud data is divided into a missing region and a non-missing region, the missing region sorting means, and the missing region located at the extreme end on one side in the predetermined cross-sectional direction toward the other side in the predetermined cross-sectional direction, Lack of In the region or the non-missing region sequentially combined by merging missing area, merging the a predetermined cross-sectional direction of the width of the non-missing region to be merged, the accumulated value of a predetermined cross-sectional direction of the width of said missing region is merged with wherein a difference between the accumulated value of a predetermined cross-sectional direction of the width of the non-missing region, and are missing region merging means to terminate sequentially merging the missing area or the non-missing region based on being, the merged missing region Perforated region determining means for determining the merged missing region as a perforated region based on the width in the predetermined cross-sectional direction.

The welding bead defect detection method for detecting defects in a weld bead according to claim 2, wherein point cloud data in a predetermined cross-sectional direction is acquired for the surface shape of the weld bead, and the point cloud data and the point cloud data are smoothed The point cloud data as an abnormal point, based on a comparison with a smoothed model, or a comparison with a broken line model in which the surface shape of the weld bead is approximated to a broken line using the smooth model. Sorting into normal points, and classifying the point cloud data into missing regions and non-missing regions based on the presence or absence of the abnormal points between the adjacent normal points and the distance between the adjacent normal points When the missing region located at the extreme end on one side in the predetermined cross-sectional direction is merged sequentially with the missing region or the non-missing region toward the other side in the predetermined cross-sectional direction to form a merged missing region. , The difference between the and the predetermined cross-sectional direction of the width of the non-missing region, the accumulated value of a predetermined cross-sectional direction of the width of the non-missing region is merged with the accumulated value of a predetermined cross-sectional direction of the width of said missing regions are merged to be engaged And ending the sequential merge of the missing region or the non-missing region , and determining the merged missing region as a perforated region based on the width of the merged missing region in a predetermined cross-sectional direction. is there.

  According to the weld bead defect detection device and the weld bead defect detection method of the present invention, the detection accuracy of the weld bead defect can be improved.

The schematic diagram which shows the surrounding structure of a weld bead defect detection apparatus. The block diagram which shows the structure of a weld bead defect detection apparatus. The flowchart which shows the flow of a weld bead defect detection process. The flowchart which shows the flow of a point cloud data acquisition process. The graph which shows a master model. The graph which shows a master model and a welding model. The graph which shows a master model, a welding model, and a Z direction difference. The flowchart which shows the flow of a point cloud data classification process. The graph figure which shows a master model, a welding model, and a smooth model. The graph figure which shows a master model, a smooth model, and a broken line model. The graph which shows a missing area and a non-missing area. The flowchart which shows the flow of a missing area merge process. The graph which shows cumulative value (beta). Another graph showing the accumulated value β. Another graph showing the accumulated value β.

The surrounding configuration of the weld bead defect detection device 10 will be described with reference to FIG.
In FIG. 1, a broken line indicates an electric communication line, and a two-dot chain line indicates a laser beam. In the following, description will be made according to the X, Y, and Z directions shown in FIG.

  The weld bead defect detection device 10 is an embodiment of a weld bead defect detection device according to the present invention. The weld bead defect detection device 10 is connected to a surface shape measurement device 15 that measures the surface shape of the welded part 20.

  The welded part 20 is obtained by partially overlapping the lower plate 21 and the upper plate 22 and welding the overlapped portions. The lower plate 21 and the upper plate 22 are gently curved. A portion where the lower plate 21 and the upper plate 22 are partially overlapped (welded portion) is a weld bead 25.

  The surface shape measuring device 15 is a laser scanner, and measures the shape of the cross section in the X direction substantially orthogonal to the weld bead 25 extending in the Y direction with respect to the surface shape of the welded part 20. More specifically, the surface shape measuring device 15 irradiates a linear laser beam that traverses the weld bead 25 of the welded part 20 by an optical cutting method, and projects the region (the welded part 20 of the welded part 20). Take the two-dot chain line). The surface shape measuring device 15 measures the surface shape of the welded part 20 at regular intervals along the Y direction in which the weld bead 25 extends.

  The weld bead defect detection device 10 is configured by a general computer, and includes a storage device (not shown) and an arithmetic device (not shown). The surface shape data measured by the surface shape measuring device 15 is input to the weld bead defect detecting device 10.

The configuration of the weld bead defect detection device 10 will be described with reference to FIG.
The weld bead defect detection device 10 includes a point cloud data acquisition unit 100, a point cloud data classification unit 200, a missing region classification unit 300, a missing region merging unit 400, and a perforated region determination unit 500. Yes.

The configuration of the weld bead defect detection step S10 in the weld bead defect detection method will be described with reference to FIG.
The weld bead defect detection step S10 is an embodiment of the weld bead defect detection step in the weld bead defect detection method according to the present invention. The weld bead defect detection step S10 includes a point cloud data acquisition step S100, a point cloud data sorting step S200, a missing region sorting step S300, a missing region merging step S400, and a perforated region determining step S500. Yes.

  In addition, each process S100-S500 of welding bead defect detection process S10 is a process performed by each means 100-500 of the weld bead defect detection apparatus 10, respectively. More specifically, the point cloud data acquisition step S100 is performed by the point cloud data acquisition unit 100, the point cloud data classification step S200 is performed by the point cloud data classification unit 200, and the missing region classification step S300 is omitted. The missing region merging step S400 is performed by the missing region merging unit 400, and the perforated region determining step S500 is performed by the perforated determining unit 500.

The point cloud data acquisition step S100 will be described with reference to FIGS.
The point cloud data acquisition step S100 is a step of acquiring the cross-sectional shape in the X direction as the point cloud data for the surface shape of the weld bead 25 and specifying the bead section W where the weld bead 25 exists in the point cloud data.

The flow of the point cloud data acquisition step S100 will be described with reference to FIG.
The point cloud data acquisition step S100 includes a master shape acquisition step S110, a weld shape acquisition step S120, an alignment step S130, and a bead section specifying step S140.

Master shape acquisition process S110 is demonstrated using FIG.
In FIG. 5, the horizontal axis indicates the X direction (cross-sectional direction in which the surface shape of the welded part 20 is measured), and the vertical axis indicates the Z direction (vertical direction). In FIG. 5, the solid line indicates the master model ML.

  In master shape acquisition process S110, master model ML is acquired as a cross-sectional shape of the X direction about the surface shape of the welding component 20 before welding. The master model ML includes a lower plate master model MUL that is the surface shape of the lower plate 21 and an upper plate master model MOL that is the surface shape of the upper plate 22. The master model ML is acquired from the product CAD data of the lower plate 21 and the upper plate 22, respectively.

  In the master shape acquisition step S <b> 110, the coordinate position is determined with the point where the end of the back surface of the upper plate 22 is in contact with the surface of the lower plate 21 as the reference point CU in the X direction cross section. That is, the inner boundary point between the upper plate 22 and the lower plate 21 is set as the reference point CU. Further, in the X-direction cross section, the coordinate positions are determined with the sections of the surfaces of the lower plate 21 and the upper plate 22 that are not affected by welding and sufficiently separated from the weld bead 25 as the respective fitting sections RU / RO. To do.

The welding shape acquisition process S120 will be described.
In the welding shape acquisition step S120, the surface shape of the weld bead 25 measured by the surface shape measuring device 15 is acquired as a welding model BL as point cloud data of the cross-sectional shape in the X direction.

The alignment step S130 will be described with reference to FIG.
In FIG. 6, the horizontal axis indicates the X direction (the cross-sectional direction in which the surface shape of the welded part 20 is measured), and the vertical axis indicates the Z direction (vertical direction). In FIG. 6, the solid line indicates the welding model BL, and the broken line indicates the master model ML.

  In the alignment step S130, the master model ML is superimposed on the welding model BL. At this time, the lower plate master model MUL and the upper plate master model MOL are superimposed on the welding model BL independently. More specifically, the lower plate master model MUL is overlaid on the welding model BL in the fitting section RU, and the upper plate master model MOL is overlaid on the welding model BL in the fitting section RO.

  In the alignment step S130, the reference point CU on the lower plate master model MUL side and the reference point CU on the upper plate master model MOL side are welded models when the master model ML is superimposed on the welding model BL. The distortion caused by welding in BL affects the reference points CU ′ and CU ′, and their positions do not match. Therefore, an intersection where the surface of the lower plate master model MUL superimposed on the welding model BL and the end surface of the upper plate master model MOL are extended to each other is defined as a reference point CU.

The bead section specifying step S140 will be described with reference to FIG.
In FIG. 7, the horizontal axis indicates the X direction (the cross-sectional direction in which the surface shape of the welded part 20 is measured), and the vertical axis indicates the Z direction (vertical direction) and the Z direction difference ΔZ. In FIG. 7, in the upper graph, the solid line indicates the welding model BL, the broken line indicates the master model ML, and in the lower graph, the alternate long and short dash line indicates the Z difference line ΔZL.

  The bead section specifying step S140 divides the welding model BL into the area G based on the difference between the welding model BL and the master model ML, and further merges the divided areas G so that the bead section W where the weld bead exists. It is a process of specifying.

  In the bead section specifying step S140, first, a Z direction difference ΔZ that is a difference in the Z direction between the welding model BL and the master model ML is calculated and set as a Z difference line ΔZL. Next, the Z difference line ΔZL is divided into regions G having different properties in the X direction.

  More specifically, the Z-direction difference ΔZ and the threshold value TH11 from the start point R1 that is a boundary point near the weld bead 25 in the fitting section RU to the end point R2 that is a boundary near the weld bead 25 in the fitting section RO. And compare. A region G in which the absolute value of the Z-direction difference ΔZ is equal to or less than the threshold value TH11 is set to 0 group (for example, the region G1 or the region G3). Further, a region G in which the Z direction difference ΔZ is larger than the threshold value TH11 is set as a + group (for example, a region G4 or a region G8). A region G in which the Z direction difference ΔZ is smaller than the threshold value −TH11 is defined as a group (for example, a region G2 and a region G6).

  Next, of the divided regions G, the region G having the largest absolute value of the area (integrated value of the Z difference line ΔZL) is selected from the + group and the − group and is designated as the maximum region G *. When the + group or the − group does not exist, it is assumed that the bead section W does not exist.

  Next, for the adjacent region G from the maximum region G * toward both sides in the X direction, the region G has an absolute value of the area of the region G (the integrated value of the Z difference line ΔZL in the X direction) equal to or greater than the threshold value TH12. In the case where the distance hG between the boundary points of the welding model BL is equal to or less than the threshold value TH13, or the region G is 0 group, the adjacent regions G are sequentially merged.

  The adjacent areas G are sequentially merged from the maximum area G * toward both sides in the X direction, and the area where the merge is completed is defined as a bead section W. In FIG. 7, the region G4 is the maximum region G *, the region G2, the region G3, the region G5, the region G6, the region G7, and the region G8 are “the absolute value of the area is greater than or equal to the threshold value TH12”, “the distance hG is It corresponds to either “threshold value TH13 or less” or “0 group”. On the other hand, since the absolute value of the area of the region G1 is smaller than the threshold TH12 and the distance hG between the boundary points of the welding model BL in the region G9 is larger than the threshold TH13, the bead section W is a region from the region G2 to the region G8.

The point cloud data classification step S200 will be described with reference to FIGS.
In the point cloud data classification step S200, the welding model BL as the point cloud data is compared with the smooth model SL obtained by smoothing the welding model BL, or the surface shape of the smooth model SL and the weld bead 25 is approximated by a broken line. This is a step of separating the welding model BL (point cloud data) into an abnormal point and a normal point based on the comparison with the model VL.

  In the point cloud data classification step S200, first, a smooth model SL obtained by smoothing the welding model BL by moving average is created. And point cloud data which forms welding model BL is set to point Qj (j = 1, ... J). Further, the point cloud data forming the smooth model SL is defined as a smooth point Sj (j = 1,... J). Note that the point where j = 1 is the starting point R1. The point where j = J is the end point R2 (see FIG. 9).

  In step S210, it is confirmed whether the point Qj and the smooth point Sj exist in the bead section W. When the point Qj and the smooth point Sj exist in the bead section W, the process proceeds to step S220. Otherwise, the process proceeds to step S211.

  In step S211, it is confirmed whether or not the shortest distance from the smooth point Sj to the master model ML is larger than the threshold value TH24. When the shortest distance is larger than the threshold value TH24, the point Qj and the smooth point Sj are determined as abnormal points. In other cases, the point Qj and the smooth point Sj are determined as normal points.

  In this way, in step S211, it is determined whether the point Qj and the smooth point Sj that are not in the bead section W are abnormal points or normal points based only on the distance from the master model ML.

Steps S220 and S230 will be described with reference to FIG.
In FIG. 9, the horizontal axis indicates the X direction (cross-sectional direction for measuring the surface shape of the welded part 20), and the vertical axis indicates the Z direction (vertical direction). In FIG. 9, the solid line indicates the welding model BL, the broken line indicates the master model ML, and the alternate long and short dash line indicates the smooth model SL.

  In step S220, the shortest distance h (Qj) from the point Qj to the master model ML and the shortest distance h (Sj) from the smooth point Sj to the master model ML are calculated. Next, it is confirmed whether the shortest distance h (Qj) is larger than the threshold value TH21 or whether the shortest distance h (Sj) is larger than the threshold value TH21. When the shortest distance h (Qj) is larger than the threshold value TH21 or when the shortest distance h (Sj) is larger than the threshold value TH21, the process proceeds to step S230. In other cases, the point Qj and the smooth point Sj are determined as normal points.

  In this way, in step S220, the point Qj and the smooth point Sj that are close to the master model ML are determined as normal points.

  In step S230, a Z component distance dZ that is a distance in the Z component between the point Qj and the smooth point Sj is calculated. And it is confirmed whether Z component distance dZ is larger than threshold value TH22. When the Z component distance dZ is larger than the threshold value TH22, the point Qj and the smooth point Sj are determined as abnormal points. Otherwise, the process proceeds to step S240.

  Thus, in step S230, the welding model BL and the smooth model SL are compared, and the point Qj and the smooth point Sj having a large difference are determined as abnormal points.

Step S240 and step S241 will be described with reference to FIG.
In FIG. 10, the horizontal axis indicates the X direction (cross-sectional direction for measuring the surface shape of the welded part 20), and the vertical axis indicates the Z direction (vertical direction). In FIG. 9, the solid line indicates the broken line model VL, the broken line indicates the master model ML, and the alternate long and short dash line indicates the smooth model SL.

  In step S240, first, a broken line model VL is created in which the end points W1 and W2 of the bead section W and the smooth point Sj are connected by a straight line. Next, the shortest distance hj (Sk) from the smooth point Sk (k = j1,..., J2) in the bead section W to the broken line model VL is calculated separately.

  In step S241, it is confirmed whether or not the majority (50% or more) of the shortest distance hj (Sk) greater than the threshold value TH23 among the smooth points Sk (k = j1,..., J2) in the bead section W exists. When the shortest distance hj (Sk) greater than the threshold value TH23 exists, the point Qj and the smooth point Sj are determined as abnormal points. In other cases, the point Qj and the smooth point Sj are determined as normal points.

  In this way, in step S240, it is determined whether the smooth point Sj is close to the welding model BL.

The missing region sorting step S300 will be described with reference to FIG.
In FIG. 11, the horizontal axis indicates the X direction (cross-sectional direction for measuring the surface shape of the welded part 20), and the vertical axis indicates the Z direction (vertical direction). In FIG. 9, the solid line indicates the welding model BL, and the broken line indicates the master model ML.

  In the missing region sorting step S300, based on the abnormal point and the distance between adjacent normal points, the welding model BL is divided into a missing region GGm (m = 1,...) And a non-missing region GNm (m = 1). ,...).

  In the missing area separation step S300, between the start point R1 and the end point R2, two adjacent normal points have a distance between two adjacent normal points that is greater than the threshold TH31 or between two adjacent normal points. If there is an abnormal point, a section having two normal points adjacent to each other is defined as a missing region GGm (m = 1,...). Further, a section sandwiched between the missing areas GG is defined as a non-missing area GNm (m = 1,...).

The missing region merging step S400 and the perforated region determining step S500 will be described with reference to FIGS.
In FIG. 12, the missing region merging step S400 is shown on the upper side of the drawing, and the perforated region determining step S500 is shown on the lower side of the drawing.

  The missing region merging step S400 (steps S410 to S480) is missing from the start point R1 to the end point R2 toward the plus side in the X direction with respect to the missing region GG1 located at the extreme end on the minus side in the X direction. When the region GGm (m = 1,...) Or the non-missing region GNm (m = 1,...) Is sequentially merged into the merged missing region GGG, the width in the X direction of the merged non-missing region GNm DNm (m = 1,...) And the cumulative value of the merged missing region GGm in the X direction width DGm (m = 1,...) And the non-missing region GNm merged in the X direction width DNm (m = 1,...) And a cumulative value β that is a difference from the cumulative value.

  The perforated region determination step S500 (steps S510 to S530) is a step of determining the merged missing region GGG as a perforated region based on the width in the X direction of the merged missing region GGG (accumulated maximum value αmax described later). .

The flow of the missing region merging step S400 will be described using the upper side of FIG.
In the missing region merging step S400, the missing region GGm (m = 1,... M) is merged from the start point R1 to the end point R2. Further, in the following, it is assumed that the cumulative value α, the cumulative value β, the cumulative maximum value αmax and the cumulative maximum value βmax, which are the respective maximum values when merging, are calculated as they are merged.

  In step S410, among the remaining missing areas GGm, the missing area GGm at the extreme end on the minus side in the X direction is set as the starting missing area GGm0. It is assumed that the missing region GGm and the non-missing region GNm are sequentially merged from the starting missing region GG0 toward the plus side in the X direction.

  In step S420, the cumulative value α is initialized to 0 at the start of merging. The cumulative value β is initialized to 0 at the start of merging. The cumulative maximum value αmax is initialized to 0 at the start of merging. The cumulative maximum value βmax is initialized to 0 at the start of merging.

  In step S430, when the missing area GGm is merged, the cumulative value α is increased by the width DGm, and the cumulative value β is increased by the width DGm.

  In step S440, it is confirmed whether or not the current accumulated value β is larger than the accumulated maximum value βmax. If the current accumulated value β is larger than the accumulated maximum value βmax, the process proceeds to step S450. In cases other than that described here, process flow proceeds to Step S460.

  In step S450, the cumulative maximum value αmax is set as the current cumulative value α. The cumulative maximum value βmax is set as the current cumulative value β. Further, the current missing area GGm is set as the β maximum missing area GGmmax in which the cumulative maximum value βmax exists.

  In step S460, it is confirmed whether the missing area GGm to be merged is the missing area GGM (m = M) closest to the end point R2. When the missing area GGm to be merged is the missing area GGM (for example, when the missing area GG4 corresponds to the last missing area GGM as shown in FIG. 13), the missing areas GGm0 to β maximum missing area GGmmax are merged. The merged missing area GGG. In cases other than that described here, process flow proceeds to Step S470.

  In step S470, it is confirmed whether the non-missing region GNm having a width DNm larger than the threshold TH41 is reached as the merge is performed, or whether the accumulated value β is smaller than 0 when the non-missing region GNm is merged. As merged, the non-missing region GNm having a width DNm larger than the threshold TH41 is reached (for example, as shown in FIG. 14, when the width DN2 of the non-missing region GN2 is larger than the threshold TH41), or the non-missing region GNm is merged Then, when the cumulative value β is smaller than 0 (for example, as shown in FIG. 15, when the cumulative value β is smaller than 0 when the non-missing region GN3 is merged), the missing region GGm0 to the β maximum missing region GGmmax. Are merged into a merged missing area GGG. Otherwise, the process proceeds to step S480.

  In step S480, when merging the non-missing region GNm, the cumulative value α is increased by the width DGm, and the cumulative value β is decreased by the width DNm. Then, the process proceeds to step S430.

The flow of the perforated region determination step S500 will be described using the lower side of FIG.
In step S510, it is confirmed whether or not the cumulative maximum value αmax in the merged missing region GGG, that is, the width in the X direction of the merged missing region GGG is greater than the threshold value TH51. When the cumulative maximum value αmax is larger than the threshold value TH51, the process proceeds to step S520. Otherwise, the process proceeds to step S530.

  In step S520, the merged lack region GGG is determined as a perforated region.

  In step S530, the merged lack region GGG is not determined as a perforated region.

  Even after the perforated region determination step S500 is completed, the missing region merging step S400 is resumed from the β maximum missing region GGmmax until the end point R2 (m = M) is reached.

The effect of the weld bead defect detection method S10 will be described.
According to the weld bead defect detection method S10, the detection accuracy of the weld bead defect can be improved. That is, according to the weld bead defect detection method S10, it is possible to prevent a hole from being missed and to minimize the erroneous detection.

DESCRIPTION OF SYMBOLS 10 Weld bead detection apparatus 15 Surface shape measuring apparatus 20 Welded part 21 Lower board 22 Upper board 25 Weld bead 100 Point cloud data acquisition means 200 Point cloud data classification means 300 Missing area classification means 400 Missing area merging means 500 Perforated judgment means S10 Weld bead detection method S100 Point cloud data acquisition process S200 Point cloud data classification process S300 Missing area classification process S400 Missing area merging process S500 Perforated judgment process BL Welding model ML Master model SL Smooth model VL Polygonal line model GG Missing area GN Non-missing area DG width (missing area)
DN width (non-missing area)
GGG merged missing area

Claims (2)

A weld bead defect detection device for detecting defects in a weld bead,
Point cloud data acquisition means for acquiring point cloud data in a predetermined cross-sectional direction for the surface shape of the weld bead;
Based on a comparison between the point cloud data and a smooth model obtained by smoothing the point cloud data, or a comparison between the smooth model and a broken line model in which the surface shape of the weld bead is approximated to a broken line using the smooth model. Point cloud data classification means for classifying the point cloud data into abnormal points and normal points;
Based on the presence or absence of the abnormal point between the adjacent normal points and the distance between the adjacent normal points, the point cloud data is classified into a missing region and a non-missing region,
When the missing region or the non-missing region is sequentially merged toward the other side in the predetermined cross-sectional direction with respect to the missing region located at the extreme end on one side in the predetermined cross-sectional direction, the merged region is merged. and a predetermined cross-sectional direction of the width of the non-missing region is, the difference between the accumulated value of a predetermined cross-sectional direction of the width of the non-missing region is merged with the accumulated value of a predetermined cross-sectional direction of the width of said missing regions are merged , Based on the missing area merging means for ending the sequential merging of the missing area or the non-missing area ;
A perforated region determination means for determining the merged missing region as a perforated region based on a width in a predetermined cross-sectional direction of the merged missing region;
Comprising
Weld bead defect detection device. A weld bead defect detection method for detecting defects in a weld bead,
Obtain point cloud data in the predetermined cross-section direction for the surface shape of the weld bead,
Based on a comparison between the point cloud data and a smooth model obtained by smoothing the point cloud data, or a comparison between the smooth model and a broken line model in which the surface shape of the weld bead is approximated to a broken line using the smooth model. And classifying the point cloud data into abnormal points and normal points,
Based on the presence or absence of the abnormal point between the adjacent normal points and the distance between the adjacent normal points, the point cloud data is classified into a missing region and a non-missing region,
When the missing region or the non-missing region is sequentially merged toward the other side in the predetermined cross-sectional direction with respect to the missing region located at the extreme end on one side in the predetermined cross-sectional direction, the merged region is merged. and a predetermined cross-sectional direction of the width of the non-missing region is, the difference between the accumulated value of a predetermined cross-sectional direction of the width of the non-missing region is merged with the accumulated value of a predetermined cross-sectional direction of the width of said missing regions are merged Based on the above , finish merging sequentially the missing area or the non-missing area ,
Based on the width in the predetermined cross-sectional direction of the merged missing region, determine the merged missing region as a perforated region,
Weld bead defect detection method.

Images