Generalized Gaussian smoothing for baseline-free debonding assessment of sandwich panels

The present work proposes a baseline-free method to assess debonding on composite panels. The method is based on regressive Gaussian processes (GP) used to obtain vibration modes free of noise. A Bayesian scheme is implemented to allow the automatic determination of the model hyperparameters. The estimation of the undamaged vibration modes is carried out by modifying the optimal hyperparameters determined employing the damaged plate. The curvatures associated with the vibration modes are calculated as the second analytical derivative of the kernel function of the GP; thus, the use of numerical methods is avoided. A damage index is obtained comparing the differences between the curvatures of the damaged and undamaged modes. Finally, using k-means clustering, the damage size and location could be identified. The effectiveness of the proposed method is verified by means of numerical simulations and experimental tests. The predicted damage is compared with the actual damage in terms of size, location, and intersection. Both numerical simulations and experimental tests offer favorable results.

Automated summary

A baseline-free method utilizing regressive Gaussian processes and Bayesian scheme was proposed to evaluate debonding on composite panels. The estimation of undamaged vibration modes was carried out by modifying optimal hyperparameters determined from the damaged plate, and a damage index was obtained by comparing the differences between curvatures of damaged and undamaged modes. Finally, k-means clustering was used to identify damage size and location, showing favorable results in numerical simulations and experimental tests.