Vibrational detection of delamination in composites using a combined finite element analysis and machine learning approach.

One common challenge of using composite materials is detecting delamination failure in a non-destructive and cost-effective way. Past studies have proven the feasibility of using vibrational measurements to detect damage but have not explored the full capabilities and limitations of vibrational testing. Here, we use a finite element model of a composite plate to characterize the natural frequency, mode shape, and mode curvature tests for a variety of delamination scenarios. We find that the mode curvature test is resource-intensive to conduct but provides the best resolution in both identifying and localizing delamination. On the other hand, the natural frequency test is simple and inexpensive to conduct but can only reliably identify the presence of delamination. Additionally, a machine learning model is implemented to augment the natural frequency test, allowing both localization of damage and quantification of its severity with only the natural frequencies of modes 1–6. We are able to interpret our model and discover a phase transition for natural frequencies with different sized delaminations. This testing framework allows rapid non-destructive analysis for the iterative design of composites, accelerating the development of novel delamination-resistant materials. [ABSTRACT FROM AUTHOR]