Damage detection in multi-wire strands by discrete Wavelet-based outlier analysisand embedded guided ultrasonic waves

This paper describes a method based on novelty detection and Discrete Wavelet Transform for damage detection. A statistical discordancy test is used to determine the presence of outliers in a set of features extracted from the Discrete Wavelet Transform of ultrasonic guided-wave signals; the outliers indicate a faulty condition of the structure under investigation. The proposed approach, which is general, is applied to the quantitative detection of notch-like defects in steel strands used as cable-stays or prestressing tendons. The effect of the number of features and the effect of the ultrasonic noise level on the sensitivity of the method to the presence and the size of the defects are discussed