This study examines the possibility of fatigue crack detection using vibration data obtained from a sign pole in laboratory experiments. The fatigue cracks were introduced to the base of the pole by cyclic loading. Anomaly indicator proposed by Goi et al is applied to the acceleration measurement. The proposed method quantifies changes in modal properties of the pole by means of Bayesian hypothesis testing. Comparison with power spectrum density shows that the anomaly indicator detects fatigue cracks when its natural frequency decreased by more than 1% and when the total crack length exceeded 10% of the steel pipe circumference.
