Online damage evaluation of hole-edge crack using guided wave based Hidden Markov Model

Structural health monitoring based on guided wave plays an important role in the damage evaluation of practical application. However, the damage evaluation under time-varying environments which introduces undesired uncertainties to guided wave features is difficult to achieve reliably. In this paper, an approach of guided wave based Hidden Markov Model (HMM) method is proposed to improve the reliability of damage evaluation under timevarying environment. With this method, a left-right continuous HMM which is composed of several hidden states is trained firstly based on the time-varying affected guided wave features of different damage states. Each hidden state of HMM represents a damage state of a monitored structure under time-varying environment. Then a maximum average posterior probability based on the HMM can be obtained to evaluate the damage when a new guided wave feature is obtained during an on-line damage process. Finally, the method performance is validated by monitoring the hole-edge crack of an aluminum tensile specimen under fatigue load condition and result shows that the reliability of damage state evaluation is improved.