A two-step method for delamination detection in composite laminates using experience-based learning algorithm

Delamination in composite laminates reduces the structural stiffness and thus causes changes in the vibration responses of the laminates. Therefore, it is feasible to employ dynamic characteristics (such as natural frequencies and mode shapes) for delamination detection by using an optimization method. In the present study, a two-step method is proposed for the delamination detection in composite laminates using an experience-based learning algorithm. In the first step, one-dimensional equivalent through-thickness beam elements are employed to model the composite laminated beam and potential delamination locations are identified. In the second step, a typical three-dimensional finite mesh is utilized for the beam's modeling and the detailed delamination information (including the delamination location, size, and interface layer) is detected. This two-step method combines the advantages of the two different modeling techniques and is able to significantly reduce the computational cost without reducing detection accuracy. The proposed method is applied for an eight-layer quasi-isotropic symmetric (0/-45/45/90)(s) composited beam with different delamination situations to verify its effectiveness and robustness. The performance of the two-step method is demonstrated by comparing with the one-step method and other three state-of-the-art algorithms (CMFOA, PSO, and SSA). Moreover, the influence of artificial noise on the accuracy of the detection performance is also investigated. Both numerical and experimental results confirm the superiority of the proposed method for delamination detection in composite laminates especially for the prediction of delamination interface.