Damage Quantification in Composite Structures Using Autoregressive Models

When small damage is detected in its initial stage in a real structure, it is necessary to decide if the user must repair immediately or keep on safely monitoring it. Regarding the second choice, the present paper proposes a methodology for damage severity quantification of delamination extension in composite structures based on a data-driven strategy using autoregressive modeling approach for Lamb wave propagation. A pair of features is used based on the autoregressive (AR) model coefficients and residuals and a machine learning algorithm with Mahalanobis Squared Distance for outlier detection. The damage severity quantification is proposed through an experimentally identified smoothing spline trend curve between the damage index and its severity. The application of the methodology is demonstrated in a composite plate with various progressive damage scenarios. The proposed method proved to be able to identify and predict the localization and the damage index related to its respective extension of minimal simulated damage with promising accuracy.