TY  - BOOK
AU  - Masud, Manzar 
AU  - Supervisor : Dr. Aamir Mubashar
TI  - Investigation of Bio-Hybrid Fiber Reinforced Composites Under Impact Loading
U1  - 621 
PY  - 2025///
CY  - Islamabad : 
PB  - SMME- NUST; 
KW  - PhD in Mechanical Engineering
N1  - The integration of natural and synthetic fibers in bio-hybrid fiber-reinforced polymer
(HFRP) composites is gaining prominence in high-performance industries such as
aerospace and automotive, driven by the demand for materials that balance mechanical
performance, sustainability, and cost-effectiveness. This research adopts a dual approach,
combining experimental testing with machine learning (ML) to investigate and optimize
the mechanical performance of five composite laminates, including a pure carbon laminate
and four carbon–flax HFRP configurations with symmetric and asymmetric stacking
sequences. All laminates were evaluated through uniaxial tensile, compressive, lowvelocity impact (LVI) at energies from 30 to 75 J, and compression-after-impact (CAI)
testing. The symmetric BH3 layup, with evenly distributed flax layers, demonstrated
superior performance with only a 9% reduction in tensile strength compared to the carbon
baseline while showing a 37.71% increase in failure strain, indicating enhanced energy
absorption. Under compression, BH3 retained 86% of the carbon laminate’s strength and
81% of its modulus. In impact resistance, BH3 withstood energies up to 75 J, surpassing
the carbon configuration. To evaluate performance and economic trade-offs, two indices
were introduced i.e., the Impact Performance Index (IPI) and the Cost-Effectiveness Index
(CEI). BH3 achieved the highest impact performance and a CEI comparable to that of the
carbon laminate. Complementing the experimental work, an ML framework was employed
using stacking sequence and impact energy as inputs, and peak impact force, damage area,
and damage extension as outputs. Six algorithms were assessed, including decision tree
(DT), random forest (RF), deep neural networks (DNN) with Adam and stochastic gradient
descent (SGD) optimizers, and recurrent neural networks (RNN) with the same optimizers.
The DT model with depth 8 and 28 leaf nodes performed best for peak force prediction,
while the model with depth 6 and 23 leaf nodes was most accurate for damage area. An
RNN with SGD and four hidden layers containing 70 neurons achieved the highest
accuracy for damage extension. This integrated methodology demonstrates the potential of
HFRP laminates to deliver high mechanical performance, improved damage tolerance, and
enhanced sustainability for structural and impact-critical applications across automotive,
aerospace, sporting, and construction sectors
UR  - http://10.250.8.41:8080/xmlui/handle/123456789/54809
ER  -