Machine learning/finite element analysis - A collaborative approach for predicting the axial impact response of adhesively bonded joints with unique sandwich composite adherends
Created by W.Langdon from
gp-bibliography.bib Revision:1.8098
- @Article{MOTTAGHIAN:2023:compscitech,
-
author = "Fatemeh Mottaghian and Farid Taheri",
-
title = "Machine learning/finite element analysis - A
collaborative approach for predicting the axial impact
response of adhesively bonded joints with unique
sandwich composite adherends",
-
journal = "Composites Science and Technology",
-
volume = "242",
-
pages = "110162",
-
year = "2023",
-
ISSN = "0266-3538",
-
DOI = "doi:10.1016/j.compscitech.2023.110162",
-
URL = "https://www.sciencedirect.com/science/article/pii/S0266353823002555",
-
keywords = "genetic algorithms, genetic programming, Adhesively
bonded joints, Sandwich composites, Deep neural
network, ANN, Genetic programming and algorithm, Axial
impact analysis",
-
abstract = "Despite the increasing usage of adhesively bonded
joints (ABJs) in various industries, optimization of
their bond strength in a cost-effective manner remains
a challenging task, particularly when complex loading
scenarios such as static and dynamic compressive
loadings are considered. The task becomes even more
challenging in bonded joints with sandwich composite
adherends. This study focuses on the performances of
double-strap ABJs configured by unique sandwich
composite adherends, carbon fiber-reinforced plastic
straps, and a room-cured structural epoxy resin under
axial impact loading. A Finite Element-Cohesive Zone
(FE-CZ) model is developed to simulate the response of
the joints, and its integrity is validated against the
experimental tests at three impact energy levels. The
model is used to simulate the response of various ABJ
configurations under axial impact loading, taking into
account 13 material, geometrical, and testing-related
parameters that influence joint strength, thereby
generating 410 data sets. Subsequently, three Machine
Learning (ML) models, including deep neural networks
and genetic evolution (i.e., genetic programming and
genetic algorithms) are developed and trained by the
data sets to predict the ABJs load-bearing capacity.
The ML models explore the relationship between the
design parameters and the joint's ultimate load-bearing
capacities, leading to the development of
cost-effective and accurate empirical equations, and
optimized joint configurations",
- }
Genetic Programming entries for
Fatemeh Mottaghian
Farid Taheri
Citations