Figure 1:
Dataset and labels. The labels include the location of the electrodes, scaling or not, and electrodes' diameters (A) no scaling; (B) scaling.
Figure 2:
Electrode detection and measurement model. The model is based on YOLOv8_OBB, and the diameter regression is added as the red part shows. FPN, feature pyramid network.
Figure 3:
Flowchart of the scaling severity assessment.
Table 1:
Corrected scaling severity rule
| T́/(mm) | α | Level |
|---|---|---|
| T́ < 0.1 | α < 0.9 | None I |
| α ≥ 0.9 | ||
| 0.1 ≤ T́ < 0.3 | α < 0.7 | None II |
| α ≥ 0.7 | ||
| 0.3 ≤ T́ < 0.8 | α < 0.7 | None III |
| α ≥ 0.7 | ||
| T́ ≥ 0.8 | α < 0.7 | None IV |
| α ≥ 0.7 |
Figure 4.
Training process. Precision refers to the accuracy of scale identification. Recall represents the accuracy on the fouling samples. mAPs are used to evaluate the accuracy of electrode position boxes.
Figure 5:
Example of diameter test (A) Caliper (B) X-ray (C) CNN.
Table 2:
Diameter test results of 10 samples (mm)
| Sample | Caliper | X-ray | CNN | Δ1 | Δ2 | Δ3 |
|---|---|---|---|---|---|---|
| 1 | 2.37 | 2.31 | 2.36 | 0.06 | −0.05 | −0.01 |
| 2 | 2.01 | 2.11 | 2.08 | −0.1 | 0.03 | 0.07 |
| 3 | 2.18 | 2.22 | 2.21 | −0.04 | 0.01 | 0.03 |
| 4 | 2.17 | 2.15 | 2.19 | 0.02 | −0.04 | 0.02 |
| 5 | 2.12 | 2.02 | 2 | 0.1 | 0.02 | −0.12 |
| 6 | 2.34 | 2.31 | 2.24 | 0.03 | 0.07 | −0.1 |
| 7 | 2.48 | 2.41 | 2.37 | 0.07 | 0.04 | −0.11 |
| 8 | 2.57 | 2.53 | 2.46 | 0.04 | 0.07 | −0.11 |
| 9 | 2.34 | 2.42 | 2.41 | −0.08 | 0.01 | 0.07 |
| 10 | 2.40 | 2.21 | 2.21 | 0.19 | 0 | −0.19 |
| Average | - | - | - | 0.029 | 0.016 | −0.045 |