Adaptive Gaussian mixture model based structural damage monitoring method under time-varying conditions

Abstract The low reliability of damage monitoring under Time-Varying Conditions (TVCs) is a key challenge for the practical application of Structural Health Monitoring (SHM) technology to aircraft structures. Among the existing SHM methods, Guided Wave (GW) and piezoelectric sensor based SHM technique is a promising method due to its high damage sensitivity and long monitoring range. Nevertheless the reliability problem still should be addressed. To deal with this problem, an adaptive Gaussian Mixture Model (GMM) based structural damage monitoring method is proposed. The probability distribution of GW features under TVCs is represented by the GMM. The center of each Gaussian component in a GMM is determined adaptively first. Then a global optimal GMM which could change with TVCs is constructed by the expectation maximization algorithm. Finally, along with damage monitoring process, the variation of GMMs is measured by a probabilistic similarity based damage index which allows normalized detection of structural damage. The utility of the proposed method is validated in the hole-edge crack monitoring of an aluminum tension sample under varying loading condition.