Detection and quantitative evaluation of surface defects in wire and arc additive manufacturing based on 3D point cloud

Figures & data

Figure 1. Deposition system with laser scanner.

Table 1. Line-laser scanner parameters.

Model	Scan height	Scan X length	Z-axis accuracy	X-axis accuracy	Single line points
SR7400	200mm	240mm	5μm	90μm	3200

Figure 2. Defect detection objective.

Figure 3. Processes of defect detection and evaluation system.

Figure 4. The curvature and normal vector of the point cloud; (a) 3D point cloud; (b) normal vector of point cloud; (c) curvature of the point cloud.

Figure 5. Weld bead extraction process, (a) normal vector Z-component two-dimensional graph; (b) segmentation threshold of 240; (c) closed operation; (d) extracted weld bead contour; (e) removal of noise around the contour based on contour area; (f) opening and closing operation.

Figure 6. Extraction process of weld bead centreline; (a) result of morphological thinning; (b) the simplified centreline; (c) the centreline in the 3D point cloud.

Table 2. RANSAC line fitting algorithm summary.

RANSAC Line Fitting Algorithm
1	Select two random points to determine the line
2	Solve the coefficients of the line
3	Calculate how many of the original points meet the requirement for a given tolerance and save the number of points that meet the requirement
4	Judge whether the prescriptive requirements are met. If the requirements are met, save the corresponding coefficient values. Otherwise, compare the existing points with the optimal solution and then store the best coefficient values as well as the coordinates of the data points.
5	Repeat the above steps for a given number of cycles

Figure 7. Humps defect detection; (a) image of weld bead; (b) the real height curve and RANSAC fitted line; (c) height difference; (d) curvature radius of the height curve; (e)the combined index; (f) result of hump detection.

Figure 8. Detection failure case: (a) Triangular deposition; (b) Extracted centreline and actual centreline.

Figure 9. Regional growth segmentation and filtering region.

Figure 10. Defect evaluation process; (a) extraction of the weld bead cross-section; (b) polynomial fitting; (c) surface reconstruction of the defect.

Figure 11. Test cases.

Figure 12. Result of weld extraction.

Figure 13. Result of defect detection.

Figure 14. Detection results of interlayer accumulation; (b) The results of defect detection on the second layer; (c) The results of defect detection on the fourth layer; (d)The results of defect detection on the ninth layer.

Table 3. Defect detection result.

Real type	Algorithm detection result				Summary
	Normal	Pore	Bulge	Hump
normal	–	13	9	3	25
pore	38	512	23	0	573
bulge	9	18	217	0	244
hump	4	0	0	53	57

Table 4. Time statistics of normal and curvature are calculated.

Case No.	Deposition type	Defect quantity			Point cloud scale	Computation time(s)
		Humps	Pores	Bulges
Case A	A(line)	3	13	6	147181	13.33
Case B	A(line)	1	1	7	141930	6.83
Case C	A(line)	2	8	0	151479	8.54
Case D	A(line)	4	11	5	147409	12.36
Case E	A(curve)	0	7	3	476634	11.50
Case F	A(curve)	6	9	2	487246	11.54
Case G	B(curve)	2	4	3	514610	9.54
Case H	C(line)	2	6	4	72048	8.30

Notes: Deposition type, A: Single-layer single-bead deposition; B: Multi-layer single-bead deposition; C: Multi-layer multi-bead deposition.

Figure 15. (a) After the first layer; (b) After the first bead of the second layer; (c) point cloud image of (a); (d) point cloud image of (b); (e) Overlapping of (a) and (b), blue for (a), red for (b); (f) Extracting the point cloud of current bead to be detected.

Table 5. Evaluation metrics of the algorithm.

	Accuracy	Precision	Recall	F1
Pore	–	94.3	89.4	91.8
Bulge	–	87.1	88.9	88.0
Hump	–	94.6	93.0	93.8
Summary	90.5	92.0	90.4	91.2

Figure 16. Result of defect evaluation.

Table 6. Defect evaluation result.

	Algorithm measurement			Manual measurement
	Elevation (mm)	Area (mm^2)	Volume (mm^3)	Elevation (mm)	Area (mm^2)	Volume (mm^3)
Defect 1	1.69194	28.326	16.134	1.74	22	15.2
Defect 2	3.75626	405.834	495.468	3.68	380	460
Defect 3	1.43365	154.054	47.133	1.48	135	50
Error	2.65%	16.55%	5.03%	-	-	-

Data availability statement

The data that support the findings of this study are openly available in Mendeley Data at doi: 10.17632/vm5rmhtxbx.1.