Adaptive GMM and OTWD-based structural crack quantification under random load

Abstract

A challenge for online guided wave-based structural health monitoring is the reliable crack quantitative evaluation under time varying conditions (TVCs). This paper proposed an adaptive guided wave (GW)-Gaussian mixture model (GMM) and optimal transport Wasserstein distance (OTWD)-based structural crack quantification method. Firstly, adaptive GMM tracking the probability distribution of GW signal features obtained under different structural healthy states is utilized to suppress the uncertain influence caused by TVCs. Then, OTWD is proposed to measure the GMM difference with improved linearity by finding the optimal transport between two GMMs. The OTWD between GMMs established under structural healthy state and monitoring state respectively is utilized as the damage indicator, where the data discrepancy caused by TVCs is further reduced. Finally, the quantitative relationship between OTWD and crack length is established to estimate the crack length. The overall method is demonstrated on the aircraft attachment lug specimens under random load. The generalization and necessity of the method is verified by cross validation and comparison with existing methods. The results illustrate the improved reliability and accuracy of the proposed method in crack quantitative evaluation under the uncertain influence of random load.