Machine learning prediction models for investigating vibration properties of epoxy resin under moisture conditions
File(s)
Accepted version
Author(s)
Type
Journal Article
Abstract
Epoxy resins used in engineering applications are commonly exposed to wet environment during intended service life, which causes vibration property degradation and increasing risk of structural failure. In this work, vibration properties of epoxy resin plate under different moisture conditions are predicted with various sizes and boundary conditions using developed machine learning (ML) models. The dataset of epoxy vibration is established first, where values in the dataset are calculated with five moisture contents using previously developed meshless model. The dataset from meshless simulation is used to train ML models of epoxy vibration using six different algorithms, including support vector machine, decision tree, random forest, gradient boosting decision tree, extreme gradient boosting, and artificial neural network. It is found that the prediction model developed using extreme gradient boosting algorithm shows the highest accuracy of 99.9% and strong reliability. Using this model, vibration properties of epoxy resin with a series of sizes and boundary conditions are predicted under various moisture contents from dry case to saturated case, which deepens the understanding of the effects of wet environments on the vibration responses of epoxy resins. The results could be used for analysis of durability of epoxy resin, and the developed ML prediction models contribute to investigating vibration property of epoxy resin under different moisture conditions, which is crucial for ensuring durability of epoxy resin in wet environment.
Date Issued
2024-11
Date Acceptance
2024-07-25
Citation
International Journal of Non-Linear Mechanics, 2024, 166
ISSN
0020-7462
Publisher
Elsevier
Journal / Book Title
International Journal of Non-Linear Mechanics
Volume
166
Copyright Statement
Copyright © 2024 Elsevier Ltd. This is the author’s accepted manuscript made available under a CC-BY licence in accordance with Imperial’s Research Publications Open Access policy (www.imperial.ac.uk/oa-policy)
License URL
Identifier
https://doi.org/10.1016/j.ijnonlinmec.2024.104857
Publication Status
Published
Article Number
104857
Date Publish Online
2024-07-26