Impact Detection in Complex Composite Structures using Acoustic Emission and Improved Sensing Logic

Kurzfassung

In the aerospace industry, Structural Health Monitoring (SHM) systems are essential to optimize maintenance and prevent catastrophic failures. This work aims to improve the accuracy of damage localization using the Acoustic Emissions (AE) technique, overcoming the intrinsic limitations of the ∆T method applied to composite materials (CFRP) compared to isotropic materials. The main objective is to localize the source of the damage within a target area of approximately 20×20mm, developing and comparing the ∆T method with a Machine Learning algorithm based on Random Forest.

The experimental campaign, conducted at the DLR, analyzed two test plates, one made of aluminum and one made of composite material. On aluminum, the ∆T method confirmed high accuracy with an error of 5.61 mm. On the CFRP plate, the intended target was successfully achieved in the internal area circumscribed by the sensors, where the ∆T method recorded an average error of 17.92mm and a localization radius of 20.83 mm (with 95% confidence). Evaluating the entire surface of the CFRP plate, the Random Forest algorithm achieved an accuracy of 34.17 mm and 68.19 mm of radius, outperforming the standard ∆T on the global area. The results validate the effectiveness of the methodology in the area circumscribed by the sensor network, providing a solid basis for future implementations on more complex aerospace structures.