PSI - Issue 37

Kafayat Eniola Hazzan et al. / Procedia Structural Integrity 37 (2022) 274–281 Hazzan and Pacella/ Structural Integrity Procedia 00 (2019) 000 – 000

278

5

3. Results and discussion The IM technique was applied to a range of images to evaluate the accuracy of the method. The coordinates of the cracks were aligned over the BSEM to compare the percentage of overlap pixels or pixels in common (Bergs et al., 2020). The differences between the pixels were calculated by comparing the pixel value in each position. If the value of pixel in a location is the same the difference was 0. If the values were different, the result was either 1 or -1 as described in Equation 1 and shown in Fig 4. The absolute value sum of this calculation across the image array indicates how different the crack regions are to the SEM, effectively comparing the intersection points across the image and identification method. Ψ ൌ ͳͲͲ − − ൐ ∗ ͳͲͲ ሺͳሻ

Fig 4. a) Example of crack extracted using IM from WC SEM to compare intersection and union. b) Schematic of pixel value comparison to calculate identification accuracy; the 0 values are accurate regions. Where BSEM is the binary SEM image and IM is the identification method. The IM technique was applied to 49 experiments (Appendix A) with different laser parameters described in 2. Methodology. The accuracy results ranged from 95.75 - 97.99 %. Table 1 shows a few example results, the full list of results can be found in Appendix A. The accuracy is comparable to other studies that have used image processing techniques to identify macro-cracks in concrete (Mohan and Poobal, 2018) and pavements (Ranjbar et al., 2021).

Table 1. Extract of crack identification accuracy results from laser engineered WC SEM. Fluence (J/cm 2 ) Frequency (kHz) No. of cracks

Method accuracy (%)

0.030 0.050 0.070 0.070 0.099 0.099 0.149 0.200

52.5 28.0

21

97.29 95.76 97.27 96.70 97.00 97.99 97.10 96.58

135 126

100.0

5.0

39

40.0 100 28.0

102

60 68 77

100.0

There were 3 types of failure modes: false positives, false negatives, and broken cracks (Fig 5.). Broken cracks were regions were the method identified a significant portion of the crack but failed to fully capture the ends as the cracks get fainter. Broken cracks were the most common failure mode across the results, with only 39 % results having