Using machine learning to increase efficiency in design of experiments for cyclic characterization of fibre-reinforced plastics
Editor: Dieter Krause, Kristin Paetzold-Byhain, Sandro Wartzack
Author: Marc Gadinger, Christian Witzgall, Thomas Hufnagel, Sandro Wartzack
Institution: Friedrich-Alexander-Universitat Erlangen-Nurnberg, Engineering Design, Germany
DOI number: 10.35199/dfx2023.12
Efficient characterization of fatigue behavior plays a crucial role in engineering design as it reduces the financial costs associated with expensive experimental tests. Existing methods for characterizing the fatigue behavior of fibre-reinforced plastics have proven inefficient due to the oversight of important design parameters, such as fibre orientations. To address this challenge, we propose an innovative approach based on Gaussian process regression. Our approach integrates previously unaccounted design parameters into the decision-making process, ensuring that optimal design points are selected for testing. By doing so, we maximize the gain of knowledge within the model, resulting in improved efficiency and accurate characterization of fatigue behavior.