Automated nondestructive characterization of anisotropic materials

• Applicants: Professor Dr.-Ing. Michael Kupke; Professor Dr. Markus Sause
• Subject Area: Polymeric and Biogenic Materials and Derived Composites ; Production Automation and Assembly Technology
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 570064252

The focus of the project is on developing an advanced method for determining the stiffness tensor (ST) of composite materials using ultrasonic testing and robotic systems. This innovative strategy aims to improve the accuracy and flexibility of non-destructive testing, especially for complex structures such as rocket booster pressure vessels made of carbon fiber reinforced polymer (CFRP). An important innovation is the direct use of A-scan ultrasonic raw data for training neural networks to avoid information loss and potential errors associated with conventional preprocessing steps. This approach is expected to increase the accuracy of ST determination. Additionally, the project aims to expand classification options for various layer configurations beyond basic unidirectional, cross-ply, and quasi-isotropic layers. This includes identifying stacking sequences and determining the ST for a wide range of configurations using classical laminate theory to calculate the homogenized ST. To ensure the robustness of machine learning models, training data from different sources are collected in the project, including experimental measurements and simulations using Dispersion Calculator (DC) and COMSOL Multiphysics Software. This diverse dataset is intended to improve the accuracy and generalizability of the model. Experimental validation is crucial, comparing the ST values obtained by ultrasonic testing with the results of mechanical tests and ultrasonic phase spectroscopy to ensure the reliability of the proposed method. A significant portion of the project is dedicated to automating ultrasonic testing using a robotic system. This involves developing a procedure for performing Polar Sweeps - rotational scans to detect Lamb waves from various angles - on test specimens and ultimately on full-sized components such as hydrogen pressure vessels. The research methodology includes designing and manufacturing test specimens with different layers, characterizing these materials through destructive and non-destructive methods, and developing an automated Polar Sweep method for data collection. Machine learning techniques are used to analyze the A-scan data, directly linking signal features with the ST components. Additionally, the project plans to further automate data collection and apply the developed methods to confirm the concept on full-sized components.

• DFG Programme: Research Grants
• Co-Investigator: Dr. Armin Huber