Scalable Monitoring System for the Localization of Damaging Events in Thin-Walled CFRP Structures Based on Acoustic Emission Analysis and Neural Networks

Abstract

Listen

Translate article

A constant challenge for the design and operation of CFRP primary structures is their sensitivity towards impact loading. This can lead to the formation of externally invisible delaminations which endanger the structural integrity. In practice, this circumstance is encountered with elaborate inspections or conservative design. Structural Heath Monitoring (SHM) systems offer the potential for permanent monitoring and represent an alternative approach that has drawn more attention in the last decade. The biggest barriers to market entry for this technology are system costs and reliability. This study is dedicated to these two points with the development of a low-cost system with which representative acoustic emission sources can be located reliably in a complex CFRP structure. The implementation is carried out using acoustic emission analysis, which represents a promising solution for the integral monitoring of primary structures. It is based on the detection of acoustic waves that are released during crack initiation and growth and propagate over large areas in thin-walled structures as Lamb waves. The challenges of source localization in thin-walled CFRP structures lie in the consideration of wave dispersion, anisotropic material properties, variable component geometry and interfaces. In this thesis, this complexity is captured by training a neural network. For this purpose, artificial sources are used which imitate acoustic emissions of typical damaging events in the material in frequency and mode content. The demonstrating component is an omega profile equipped with a network of piezoelectric sensors that is designed for reliable localization within a defined window. Signal processing takes place on a single-board computer which, together with a digital oscilloscope, completes the measurement chain. The system represents a modular, low-cost approach that can be transferred to other applications by adapting the hardware and training.