فرشاد محمدزاده

The propagation of guided Lamb waves has attracted significant attention from researchers in the past few decades as a highly useful non-destructive eva‎luation technique in industries such as aerospace, civil engineering, and oil and gas. Polymeric laminated composite structures based on polymer matrices have gained popularity and competed with metal structures in these industries due to their high strength-to-weight ratio, good resistance to certain environmental conditions, and cost-effectiveness in production. Despite their mentioned advantages, the difficulities in the identification of critical damages such as delaminations and the rapid progression of internal damages have made structural health monitoring of these structures imperative prior to their ultimate failure. In this study, the guided Lamb wave propagation method was investigated for detecting delamination damage in a laminated composite structure made of glass epoxy with 〖[0_2/90_6 ]〗_s layup. A virtual laboratory was established for conducting all simulations, including preprocessing, processing, and post-processing stages on both pristine and delaminated composite samples. In the preprocessing stage, dispersion curves were obtained using a semi-analytical finite element method based on the material properties of the composite. The preprocessing data will be used to validate the results of the post-processing and select the appropriate processing parameters. In the processing stage, with the aid of a Python code developed in the finite element software Abaqus, the guided Lamb wave propagation with the zeroth anti-symmetric mode was implemented in the structure with delamination placed at the interfaces between the layers and with different delamination sizes. Approximatly, 7000 unique simulations were performed. The received waves from embedded sensors in each configuration were transformed into frequency-wavenumber images using a two-dimensional fast Fourier transform. These images were then analyzed by a deep learning image classifier in three classification tasks. By eva‎luating the trained model, an accuracy of 98% was achieved for discriminating between healthy and delaminated composite samples. Another classifier accurately determined the location of the delamination within the thickness of structure with an accuracy of 90%. Furthermore, the accuracy of 90% was acheived in detecting the size of the delamination. These results indicate the capability of this model for structural health monitoring of composite structures especially in detection of delamination.

پايش سلامت سازه , گسترش امواج لمب شبيه‌سازي‌شده , كامپوزيت‌هاي پايه پليمري , روش اجزاي محدود , يادگيري عميق

Structural health monitoring , Simulated Guided Lamb waves , Polymeric composites , Finite element method , Deep learning

Farshad Mohammadzadeh

Mahmood Mehrdad Shokrieh