LAUSR

Abstract This paper aims to present a computational and experimental investigation on the performance of a new damage detection approach based on empirical wavelet transform (EWT) and an artificial neural network (ANN) named EWT-ANN to detect the damage existence, intensity, and location in a laboratory-scale model of a steel truss bridge in healthy and damaged states when exposed to white noise and impact excitations. In addition, five signal features of the decomposed mode components by EWT including energy, root mean square (RMS), shape factor, kurtosis, and entropy are extracted and defined as the targets of ANN to quantitatively and qualitatively assess the proposed approach. It is found that the proposed approach based on the time-domain signal features is efficiently able in identifying the damage location, and severity when the truss is subjected to impact loading excitation.