Uncertainty-Aware Neural Network for Data-Driven Modeling of A Foam-Damped System
L. De Malsche (TU Delft - Aerospace Engineering)
Vahid Yaghoubi – Mentor (TU Delft - Group Yaghoubi Nasrabadi)
Dimitrios Zarouchas – Graduation committee member (TU Delft - Group Zarouchas)
A. Grammatikopoulos – Graduation committee member (TU Delft - Ship and Offshore Structures)
E. Lemmens – Graduation committee member (Redwire Space N.V.)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Protecting sensitive objects during a rocket launch is imperative. For equipment going to the ISS this is done through packing them in polymer foam or bubble wrap. Simulating how well an object is packaged is computationally intensive and difficult to implement. To solve this, this thesis examines the performance of three surrogate models, LSTM, LSTM-FC-GP, and LSTM-PCA-GP, to predict the response of a foam packed object when it is subjected to a vibration. The results show that the models can accurately predict the response of the system in both the time and frequency domain. Furthermore, the LSTM models that are augmented with a GP can accurately predict the uncertainty of the system in addition to having better performance.