Colloquium: Thomas De Jonghe (ASM)
17 december 2020 09:30 - Locatie: Meeting Room 1, Faculty of Aerospace Engineering, Kluyverweg 1, Delft
Online Prognostics of Remaining Useful Properties for Cross-Ply Composites in Early Fatigue Life: A Model-Based Machine Learning Approach
High performance fiber-reinforced composite structures are becoming increasingly important in the aerospace industry. These lightweight materials offer the potential to reduce airframe weight, which in turn leads to fuel savings. However, internal damage is often created and propagated throughout the lifetime, which has a negative impact on the material properties. In order to allow for condition-based maintenance, there is a need for accurate and precise prognostics to predict future damage states. Therefore, a model-based machine learning approach is developed in this thesis to perform prognostics of the remaining useful life and properties of cross-ply composite structures in early fatigue life. To account for the multicausality and non-linearity in stiffness degradation, the crack density and delamination ratio are propagated using separate phenomenological relations that are roughly pre-trained using non-linear least squares. A particle filter then trains the phenomenological models for a specific specimen online using DIC measurements while monitoring and propagating uncertainties.
The public parts (presentation and graduation) can be accessed via: