Advanced Fracture Mechanics: Integrating Digital Image Correlation, Explainable Machine Learning, and Symbolic Regression
Please login to view abstract download link
Fatigue crack growth remains a fundamental challenge in structural integrity assessment, necessitating advanced methodologies that integrate experimental insights with data-driven approaches. In this work, we present a novel framework for fatigue crack growth experiments that leverages digital image correlation (DIC), robotics-enhanced testing, and machine learning techniques, aiming to enhance the accuracy and efficiency of crack growth characterization. Our robotic test infrastructure enables automated high-resolution DIC measurements at different length scales, significantly increasing the information-to-cost ratio compared to conventional methodologies. In addition to experimental advancements, we discover crack tip correction formulas using symbolic regression. By employing deep symbolic regression guided by physical constraints, we extract interpretable formulas that refine crack tip detection and improve the stability of crack propagation curves. The proposed method combines finite element simulations with higher-order Williams coefficients to generate physically meaningful corrections under various loading conditions. Our contribution highlights the crucial role of high-fidelity experimental data in validating and improving data-driven models. Through this presentation, we seek to engage with researchers working on neural network-based constitutive modeling and data-driven material identification, fostering interdisciplinary collaboration between experimentalists and theorists to advance the state of fatigue modeling and damage prediction.
complas_sec@cimne.upc.edu/ Telf. + 34 - 93 405 46 94
Copyright © 2024 CIMNE, All Rights Reserved. Terms of service