PSI - Issue 77

Pawel Madejski et al. / Procedia Structural Integrity 77 (2026) 357–364 Author name / Structural Integrity Procedia 00 (2026) 000–000

362

6

For the Lines pattern, porosity results from different segmentation methods show notable variation as shown in Table 2. The watershed method detected the highest total porosity of 13.72%, primarily driven by a high defect porosity of 8.99%. In contrast, the Otsu and adaptive thresholding methods yielded lower total porosity values, ranging from 7.9% to 8.6%, with more balanced contributions from structural and defect components. Region growing yielded intermediate total porosity 8.5% with lower defect porosity 3.33%. This spread suggests that watershed segmentation may overestimate defects by splitting pore regions too aggressively, whereas Otsu and adaptive thresholding provide more conservative and potentially more accurate porosity classification. The defect porosity, which is comparable to or sometimes greater than the structural porosity for the Lines pattern, suggests that manufacturing irregularities significantly impact the internal integrity, despite the intended infill design. The density verification for the Lines infill pattern exhibited a generally strong correlation between the micro CT-based estimated densities and the experimentally measured density of 1.252 g/cm³ (Table 3). Most segmentation methods, including Otsu, adaptive thresholding, and region growing, produced estimated densities within a narrow deviation range of approximately -1.4% to -0.7%, indicating a high accuracy in porosity quantification and volume reconstruction. This close agreement validates the image processing workflow’s ability to accurately differentiate between solid material and voids in the Lines pattern. The watershed method, however, showed a higher deviation (+5.3%), likely caused by its segmentation approach that tends to over-segment pore regions, leading to an overestimation of the material volume and, consequently, the density. Table 2. Structural, defect, and total porosity for Lines pattern Segmentation Method Structural porosity [%] Defect porosity [%] Total Porosity [%] Otsu 3.32 4.58 7.90 Adaptive-threshold 2.78 5.80 8.58 Watershed 4.73 8.99 13.72 Region growing 5.18 3.33 8.51

Table 3. Estimated density by segmentation method for Lines pattern Segmentation Method Experimental Density [g/cm 3 ] Estimated density [g/cm 3 ] Estimated Deviation Otsu 1.252 1.233 -1.4% Adaptive-threshold 1.252 1.242 -0.7% Watershed 1.252 1.316 +5.3% Region growing 1.252 1.241 -0.7%

3.2. Triangles Pattern

Figure 9. Originally processed CT image of Triangles pattern (slice 8).

A

B

F igure 10. A) Segmented image of Triangles pattern (slice 8) using Otsu method; and B) Overlay pore detection, green colour denotes pores.

A

B

F igure 11. A) Segmented image of Triangles pattern (slice 8) using Adaptive-threshold method; and B) Overlay pore detection, green denotes pores.