Quantification and minimization of uncertainty for guided wave-based structural health monitoring with artificial neural network approaches

The research project focuses on the development of artificial intelligence (AI)‑based Structural Health Monitoring (SHM) methods using guided ultrasonic waves. Guided ultrasonic waves enable large‑area monitoring of lightweight and fiber‑reinforced structures; however, the interpretation of the highly complex and strongly scattered signals is associated with significant uncertainties. These uncertainties arise, among other factors, from varying material properties, environmental conditions, signal processing, and measurement system influences.
Within the project, artificial neural networks, in particular feedforward and convolutional neural networks, are employed for the sequential detection, localization, and quantification of damage, as well as for the prediction of damage size. The focus is on applications to fiber‑reinforced structures, exemplarily a carbon fiber‑reinforced polymer (CFRP) plate with an omega stringer, for which data are already available via the Open Guided Waves (OGW) platform. In addition, the experimental investigations are extended by new measurements and numerical simulations in order to jointly utilize experimental and synthetic data for model development.
In particular, physical knowledge of guided wave propagation is systematically integrated into the AI models, and uncertainties arising from wave propagation, signal processing, environmental conditions, and AI modeling are quantified and minimized. Furthermore, the influence of temperature variations and surface effects, such as ice or water surface contamination, is investigated. Measurement system effects, for example due to noise or electromagnetic crosstalk, are also analyzed and statistically evaluated. By applying data and model fusion strategies, the reliability and robustness of the AI‑based SHM methods are further enhanced.

Contact Persons
Principal Investigators: Prof. Dr.-Ing. Steffen Freitag, Dr.-Ing. Jochen Moll (Universität Siegen) 
Research Assistant: M.Sc. Anastasiia Volovikova
Cooperation Partner: Dr.-Ing. Vittorio Memmolo

The project is funded by the German Research Foundation – Deutsche Forschungsgemeinschaft (DFG)