Abstract
Abstract High-speed in-line (up to 1 m s −1 ) Non-Destructive Testing requires measurement systems whose uncertainty is traceable under realistic operating conditions. In this study, the uncertainty of a directional Eddy Current Testing (ECT) sensor is quantified in accordance with Verband der Automobilindustrie (VDA) 5.1 for two defect-severity metrics: voltage-based and phase-based signal-to-noise ratios (SNR). Ten inspection speeds (0.1–1.0 m s −1 ) were investigated with ten repeated scans per speed, enabling an ANOVA-based decomposition of Type A components and a systematic evaluation of Type B contributions. Both channels exhibited a statistically significant speed dependence of comparable magnitude (between-speed (speed-effect) variance component $$u_{\textrm{AV}}\approx 1.43$$ –1.44 SNR units), while voltage-based SNR retained the lower repeatability uncertainty ( $$u_{\textrm{EV}}=1.03$$ versus 1.41 SNR units for phase). Combining Type A and Type B components yields expanded uncertainties of $$U_{\textrm{MP}}={}$$ 3.5155 (voltage) and $$U_{\textrm{MP}}={}$$ 4.0364 (phase), corresponding to minimum permissible tolerances of 7.0311 and 8.0729 SNR units, respectively. These results demonstrate that voltage-based SNR provides the most metrologically robust characteristic for downstream machine-learning analysis and establish a traceable VDA 5.1 framework for uncertainty quantification in high-speed EC inspections.