Safety evaluation of steel structures requires knowledge of corrosion progression stages. The deteriorative stage of corrosion involves multiple parameters, and thus it is difficult to be characterized by the model-based approaches. In this paper, we propose a steel corrosion stages characterization method using microwave open-ended rectangular waveguide (ORWG) probes and a statistical-based principal component analysis (PCA) method. Two ORWG probes operating in successive bands, ranging between 9.5 and 26.5 GHz, are utilized to obtain reflection coefficient spectra from specific sets of corrosion samples; i.e., uncoated corrosion progression, coated corrosion progression, and surface preparation. PCA is applied to extract corrosion progression feature from spectral responses of training samples. The robustness of the PC-based features is analyzed with influences of operating frequency, coating layer, and surface condition. It is found that the corrosion features extracted by the first principal component (PC1) from coated and uncoated corrosion samples are highly correlated to the corrosion progress regardless of probe parameter and coating layer.
11 Figures and Tables
Fig. 1. (a) ORWG probe’s aperture and field distribution in TE10 mode and (b) a model of corrosion undercoating inspected by an ORWG probe.
Fig. 10. PC1-PC3 projected values of US samples: (a) WR-62 PC1, (b) WR-62 PC2, (c) WR-62 PC3, (d) WR-42 PC1, (e) WR-62 PC2, (f) WR-62 PC3.
Fig. 2. Corrosion sample sets: (a) uncoated corrosion progression (UP), (b) coated corrosion progression (CP), and (c) uncoated surface preparation (US).
Fig. 4. PCA for feature extraction of corrosion progression.
Fig. 5. Average magnitude responses of UP samples obtained by WR-62: (a) M0, (b) M1, (c) M3, (d) M6, (e) M10, (f) M12 and WR-42: (g) M0, (h) M1, (i) M3, (j) M6, (k) M10, (l) M12.
Fig. 6. Average magnitude responses of CP samples obtained by WR-62: (a) M0, (b) M1, (c) M3, (d) M6, (e) M10, (f) M12 and WR-42: (g) M0, (h) M1, (i) M3, (j) M6, (k) M10, (l) M12.
Fig. 7. Average magnitude of the reflection coefficients of the coating layer from CP samples (C1-C12) measured by (a) WR-62 and (b) WR-42.
Fig. 8. Average magnitude responses of US samples obtained by WR-62: (a) US1, (b) US2, (c) US3, (d) US4 and WR-42: (e) US1, (f) US2, (g) US3, (h) US4.
TABLE II SURFACE PREPARATION SAMPLE’S PROPERTIES
TABLE III MEASUREMENT PARAMETERS
TABLE V STANDARD DEVIATIONS OF UNCOATED SURFACE PREPARATION (US) SAMPLES AT 0° AND 90°
Download Full PDF Version (Non-Commercial Use)