Polymer Composites

RESEARCH ARTICLE OPEN ACCESS

Investigating the Effects of Functionalized Single Wall

Carbon Nanotubes on the Cure Behavior of a Carbon/Epoxy
Prepreg System by an Optimized Parameter Approach
Murat Öz1 | Yusuf Can Uz2 | Gamze Tanoglu1 | Metin Tanoglu2 | Murat Barisik3

1Department of Mathematics, Izmir Institute of Technology, Urla, Izmir, Turkey | 2 Department of Mechanical Engineering, Izmir Institute of Technology,

Urla, Izmir, Turkey | 3Department of Mechanical Engineering, University of Tennessee at Chattanooga, Chattanooga, Tennessee, USA

Correspondence: Metin Tanoglu ([email protected])

Received: 4 September 2024 | Revised: 30 March 2025 | Accepted: 30 April 2025

Funding: This work was supported by Türkiye Bilimsel ve Teknolojik Araştırma Kurumu.

Keywords: cure behavior | functionalized SWCNTs | new algorithm | prepreg system

ABSTRACT
Carbon/Epoxy composite materials are used in a wide range of applications due to their superior performance. However, their
properties are strongly related to cross-­linking reactions occurring during the curing process, and a prior estimation of curing
parameters is the key to manufacturing the desired material. This study builds a mathematical model to solve the inverse kinetic
problem based on differential scanning calorimetry data and later presents its use in curing experiments. The method derived
(Gamze-­Murat-­Neslisah (GMN) approach) determines the pre-­exponential and activation energy of the curing process. Later,
an extended experimental study was performed. Functionalized single-­wall carbon nanotubes (F-­S WCNTs) were prepared by
oxidizing their surface with carboxyl to enhance the dispersion of the nanoparticulates. The epoxy resin systems were modified
with 0.05%, 0.1%, and 0.2% wt. F-­S WCNTs, which were impregnated on carbon fibers (CFs). The curing behavior was studied,
cure kinetic parameters were determined, and the thermal behavior was characterized. Differential scanning calorimetry (DSC)
data sets for CF/epoxy prepregs containing F-­S WCNTs were used for the verification of the proposed method. It was found that
the GMN approach is in good agreement with the experimentally measured data for all kinetic parameters. The addition of
F-­S WCNTs increased the material's curing efficiency as the CNTs enhanced heat transport in composites, reducing the activa-
tion energy. The results obtained from the GMN algorithm were also found in good agreement with the well-­known Kissinger-­
Akahira-­Sunose (KAS) and Kissinger methods, while the current GMN method revealed itself as an accurate algorithm to obtain
the activation energy.

1   |   Introduction by carbon nanotubes integrated into the resin. One significant

challenge with using epoxy resins in pre-­impregnated systems is
Hybrid composites, also referred to as multiscale, hierarchical, managing the complex reaction stages during the curing process
or nanostructured composites, incorporate reinforcements of [1]. Many studies have focused on addressing this challenge by
varying size scales, typically combining micro-­ scale contin- improving the handling and extending the shelf life of prepregs
uous fibers with nano-­scale fillers. A key advantage of these while maintaining desirable properties such as high glass tran-
composites is that they combine the benefits of traditional fiber-­ sition temperatures (Tg), high modulus, and favorable weight
reinforced materials with the enhanced functionality provided ratios [2].

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any
medium, provided the original work is properly cited and is not used for commercial purposes.
© 2025 The Author(s). Polymer Composites published by Wiley Periodicals LLC on behalf of Society of Plastics Engineers.

Polymer Composites, 2025; 0:1–16 1 of 16

https://doi.org/10.1002/pc.30039
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
well-­d ispersed carbon nanotubes significantly influences the
Summary curing of resin. Therefore, understanding the curing mecha-
nisms of carbon/epoxy materials containing CNTs is crucial
• Good dispersion of single-­wall carbon nanotubes is
obtained by functionalization. for the design, analysis, and optimization of manufacturing
processes. Recognizing prepreg processing is essential, as
• Curing of carbon fiber prepregs with carbon nano- curing conditions have a substantial impact on these proper-
tubes added epoxy is described. ties [15, 16]. Gaining insights into the effects of nano-­sized
• Effect of single-­wall carbon nanotubes on curing is materials on the curing process will facilitate the development
explained. of innovative prepregs.
• An optimized mathematical model for composite cur-
ing is developed. There has been growing interest in modeling the processing of
epoxy-­based nanocomposites to achieve high-­quality parts by
optimizing curing parameters [16, 17]. This approach can sig-
The increasing popularity of carbon nanotubes (CNTs) as nano-­ nificantly reduce the need for extensive experimental studies
scale reinforcements is largely due to their exceptional mechan- on cure cycles. The inclusion of nanofillers in the resin system
ical properties and their ability to enhance thermal conductivity can notably affect the degree of cure (α), heat of reaction (ΔH),
throughout the resin system [3–5]. Despite these advantages, and curing kinetics parameters, such as activation energy (Eα).
the use of CNTs in composites is still constrained by challenges Borchardt and Daniels [10] determined kinetic parameters using
such as poor interfacial interactions between the CNTs and the a single DSC measurement. Additionally, models such as those
polymer matrix, as well as difficulties in achieving uniform developed by Kissinger [13], Ozawa [18], Flynn–Wall–Ozawa
dispersion during processing [6–9]. Therefore, optimizing the [19], and Kissinger–Akahira–Sunose (KAS) [20] offer greater
performance of CNTs in composites depends on achieving a precision through multiple dynamic scans. Isoconversional
homogeneous and stable distribution of CNTs within the resin (model-­free) models, a type of kinetic model, provide insights
system and ensuring strong bonding between the CNTs and the into the kinetic analysis of solid-­state reactions without relying
polymer surfaces. on a specific kinetic model.

The chemical functionalization approach involves attaching Defining the complete curing of an epoxy system is challeng-
functional groups to the walls of carbon nanotubes (CNTs). This ing due to the simultaneous occurrence of multiple reactive
method is preferred over Van der Waals interactions because the processes. However, various models are useful for analyzing
covalent bonds between CNTs and epoxy offer superior stability these complex curing processes. Dubaj et al. [21] demon-
and help prevent agglomeration [10]. A common technique for strated a practical application in the non-­ isothermal cold
covalent functionalization is oxidation in acid solutions, using crystallization of polyethylene terephthalate. In their study,
strong acids such as HNO3, H2SO4, or their mixtures [11]. This the kinetic model was directly fitted to the experiments using
oxidation process can introduce defects at the open ends and/or a least squares procedure. This method processes the data
sidewalls of the CNTs. Facilitating the bonding with carboxylic without transformations, preserving the error distribution. By
(-­COOH) or hydroxyl (-­OH) groups [12] is presented as a way to using an objective function based on a maximum likelihood
control these defects. approach, the study provides reliable estimates of parameter
uncertainties.
Epoxy curing develops the cross-­l inking reactions which have
a dominant influence on the mechanical and thermal proper- Trache et al. [22] found that the TAS method is effective for ob-
ties of prepregs [13]. A three-­d imensional network develops taining all kinetic parameters in an objective manner within
through the epoxy resin. This curing process is exothermic, a reasonable computation time. This approach combines the
releasing heat. The mechanism is complex because epoxy iterative modified Coats–Redfern equation with kinetic com-
undergoes multiple physical and chemical reactions during pensation parameters (ln A = aE + b). The TAS method yields
cross-­l inking [14, 15]. These reactions are crucial for the res- consistent pre-­ exponential factors and kinetic models that
in's morphology. align with those obtained from the precise approximations
of Tang et al. Stanko et al. [23] provided a summary of the
The most commonly used method for monitoring the curing characterization of reaction kinetics for a methylene diphe-
processes is the Differential scanning calorimetry (DSC). Issues nyl diisocyanate (MDI)-­based fast curing polyurethane resin
with the quality of the final product can arise from using materi- (PUR). Their focus was on evaluating the effectiveness of iso-
als that are over-­cured or from employing an unfavorable curing conversional methods in estimating the reaction kinetics of
cycle. For example, prepregs must be stored in a cold environ- fast curing resins.
ment since the crosslinking continues even after production.
Thus, understanding the mechanisms and kinetics of curing Isoconversional methods have proven effective in analyzing
reactions is crucial. experimentally obtained DSC data, demonstrating their ability
to predict the reaction kinetics [24]. Heinze et al. [25] explored
Most research on fiber-­ reinforced epoxy composites has how to predict the curing of a polymer with an unspecified de-
primarily focused on the final properties of the material. tailed chemical composition and how to obtain the necessary
However, as optimization of processing techniques becomes material properties for these predictions. Their approach is rel-
increasingly important, there is a growing emphasis on study- atively straightforward, allowing part manufacturers to imple-
ing the curing processes of these composites. The addition of ment it without delving into the specifics of chemical reactions.

2 of 16 Polymer Composites, 2025

 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
a)

b)

c)

FIGURE 1    |    Illustration of the sample preparation steps. (a) Vacuum filtration stage of functionalized SWCNTs, (b) Drying processes of function-
alized SWCNTs, and (c) epoxy resin system dispersion of SWCNTs.

FIGURE 3    |    Non-­isothermal curing DSC curves for epoxy resin sys-

FIGURE 2    |    Drum-­t ype prepreg machine. tem at 2.5°C/min.

3 of 16
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Hosseinpour et al. [26] studied the curing behavior of an epoxy-­ iso-­conversion methods with an Optimized Parameter Approach
based resin and analyzed the curing kinetics with both model based on a tanh function nonlinear least squares fitting called
fitting and model-­free methods. While it does not provide a de- the GMN (Gamze, Murat, Neslisah) algorithm [27]. The GMN
tailed characterization, it offers sufficient accuracy for describ- method has been introduced as a rapid technique for calculating
ing the curing process. the pre-­exponential factor and energy of curing activation. It in-
tegrates a tanh function fit to measured conversion data using a
In this study, determination of kinetic parameters of CF/epoxy least squares minimization approach, along with a linear fitting
prepregs containing F-­ S WCNTs was performed by using procedure to determine the kinetic parameters.

The influence of F-­S WCNTs on the curing process and kinetic

TABLE 1    |    The heat of reaction for various carbon/epoxy prepreg
parameters of carbon fiber/epoxy-­based prepregs was studied
systems containing with/without F-­S WCNT.
using non-­isothermal differential scanning calorimetry (DSC) at
various heating rates (2.5, 5, 10, and 20°C/min). The DSC anal-
∆H (J/g) ysis under linear heating conditions revealed DSC curves with
Neat 0.05 wt% 0.1 wt% 0.2 wt%
exothermic peaks. OriginPro software was used, employing
β C/min (0.00 wt.%) F-­SWCNT F-­SWCNT F-­SWCNT the linear least squares method, to analyze these DSC curves.
From the curves, cure kinetic parameters such as the degree
2.5 43.63 97.58 146.26 113.07 of cure (α) and the reaction rate (dα/dt) were determined. The
5 46.76 97.54 150.98 110.73 activation energy of the curing reaction was calculated using
two iso-­conversional methods: the Kissinger-­A kahira-­Sunose
10 44.05 94.04 146.22 114.33
(KAS) method and the Kissinger method. The Kissinger method
20 45.86 94.17 138.08 108.16 provides an average activation energy for the entire process, en-
Average 45.07 96.08 145.38 111.57 compassing all α values, making it a quick method for param-
eter estimation. In contrast, the KAS method offers variable

FIGURE 4    |    Conversion (cure) of different carbon/epoxy prepreg systems by temperature at different F-­S WCNTs weight ratios of (a) 0.0, (b) 0.05,
(c) 0.1, and (d) 0.2.

4 of 16 Polymer Composites, 2025

 TABLE 2    |    Results of KAS method for different carbon/epoxy prepreg systems. Units: β (°C/min), Tα (°C), 1000/Tα (1/K).

0.0 wt. %F-­SWCNTs 0.05 wt. %F-­SWCNTs 0.1 wt. %F-­SWCNTs 0.2 wt. %F-­SWCNTs

α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2)

0.1 2.5 110.24 2.61 −8.49 0.1 2.5 110 2.61 −8.48 0.1 2.5 107.85 2.63 −8.45 0.1 2.5 110.49 2.61 −8.49

5 120.02 2.54 −7.97 5 117.63 2.56 −7.93 5 116.08 2.57 −7.90 5 119.82 2.55 −7.96

10 127.35 2.50 −7.39 10 128.14 2.49 −7.40 10 126.97 2.50 −7.39 10 129.84 2.48 −7.43

20 139 2.43 −6.87 20 139.9 2.42 −6.89 20 142.74 2.41 −6.93 20 143.19 2.40 −6.93

0.15 2.5 113.24 2.59 −8.54 0.15 2.5 112.96 2.59 −8.54 0.15 2.5 111.35 2.60 −8.51 0.15 2.5 114 2.58 −8.56

5 123.02 2.53 −8.02 5 121.14 2.54 −7.98 5 120.07 2.54 −7.97 5 123.82 2.52 −8.03

10 131.32 2.47 −7.45 10 131.47 2.47 −7.45 10 130.96 2.48 −7.45 10 132.9 2.46 −7.48

20 142.37 2.41 −6.92 20 144.08 2.40 −6.95 20 145.02 2.39 −6.96 20 147.15 2.38 −6.99

0.2 2.5 115.73 2.57 −8.59 0.2 2.5 115.2 2.58 −8.58 0.2 2.5 114.36 2.58 −8.56 0.2 2.5 116.49 2.57 −8.60

5 125 2.51 −8.05 5 124.12 2.52 −8.03 5 122.57 2.53 −8.01 5 126.3 2.50 −8.07

10 133.6 2.46 −7.49 10 134.4 2.45 −7.50 10 134.44 2.45 −7.50 10 135.82 2.45 −7.52

20 145.5 2.39 −6.96 20 147.46 2.38 −6.99 20 148.17 2.37 −7.00 20 149.52 2.37 −7.02

0.25 2.5 117.48 2.56 −8.62 0.25 2.5 117.2 2.56 −8.61 0.25 2.5 116.61 2.57 −8.60 0.25 2.5 118.74 2.55 −8.64

5 126.99 2.50 −8.08 5 126.12 2.51 −8.07 5 125.06 2.51 −8.05 5 128.79 2.49 −8.11

10 135.76 2.45 −7.52 10 137.17 2.44 −7.54 10 136.92 2.44 −7.54 10 138.01 2.43 −7.55

20 147.52 2.38 −6.99 20 149.84 2.36 −7.02 20 150.64 2.36 −7.03 20 151.51 2.36 −7.05

0.3 2.5 119.21 2.55 −8.65 0.3 2.5 118.94 2.55 −8.64 0.3 2.5 118.37 2.56 −8.63 0.3 2.5 120.49 2.54 −8.67

5 128.49 2.49 −8.10 5 128.11 2.49 −8.10 5 127.05 2.50 −8.08 5 130.78 2.48 −8.14

10 137.8 2.43 −7.55 10 139.14 2.43 −7.57 10 138.91 2.43 −7.57 10 140.1 2.42 −7.58

20 149.99 2.36 −7.03 20 152.14 2.35 −7.05 20 152.75 2.35 −7.06 20 153.75 2.34 −7.07

0.35 2.5 120.71 2.54 −8.67 0.35 2.5 120.44 2.54 −8.67 0.35 2.5 120.1 2.54 −8.66 0.35 2.5 122.25 2.53 −8.70

5 129.97 2.48 −8.13 5 130.1 2.48 −8.13 5 129.04 2.49 −8.11 5 132.5 2.47 −8.16

10 139.57 2.42 −7.57 10 141.12 2.41 −7.60 10 140.89 2.42 −7.59 10 138.2 2.43 −7.55

20 151.5 2.36 −7.05 20 154.27 2.34 −7.08 20 154.77 2.34 −7.09 20 155.4 2.33 −7.10

(Continues)

5 of 16
15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
 TABLE 2    |    (Continued)

6 of 16
0.0 wt. %F-­SWCNTs 0.05 wt. %F-­SWCNTs 0.1 wt. %F-­SWCNTs 0.2 wt. %F-­SWCNTs

α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2)

0.4 2.5 122.21 2.53 −8.70 0.4 2.5 121.94 2.53 −8.69 0.4 2.5 121.85 2.53 −8.69 0.4 2.5 123.73 2.52 −8.72

5 131.47 2.47 −8.15 5 131.59 2.47 −8.15 5 130.53 2.48 −8.13 5 134.24 2.46 −8.19

10 141.3 2.41 −7.60 10 143.09 2.40 −7.62 10 142.87 2.40 −7.62 10 143.73 2.40 −7.63

20 153.5 2.34 −7.07 20 156.29 2.33 −7.11 20 156.71 2.33 −7.11 20 157.18 2.32 −7.12

0.45 2.5 123.45 2.52 −8.72 0.45 2.5 123.44 2.52 −8.72 0.45 2.5 123.35 2.52 −8.71 0.45 2.5 125.22 2.51 −8.74

5 132.57 2.47 −8.16 5 133.19 2.46 −8.17 5 132.02 2.47 −8.16 5 135.73 2.45 −8.21

10 143.16 2.40 −7.63 10 144.66 2.39 −7.65 10 144.42 2.40 −7.64 10 145.15 2.39 −7.65

20 155.09 2.34 −7.09 20 158.26 2.32 −7.13 20 158.6 2.32 −7.14 20 159 2.31 −7.14

0.5 2.5 124.7 2.51 −8.74 0.5 2.5 124.67 2.51 −8.74 0.5 2.5 124.85 2.51 −8.74 0.5 2.5 126.71 2.50 −8.77

5 133.94 2.46 −8.19 5 135.05 2.45 −8.20 5 134.1 2.46 −8.19 5 137.21 2.44 −8.23

10 144.7 2.39 −7.65 10 147.02 2.38 −7.68 10 146.8 2.38 −7.68 10 146.76 2.38 −7.68

20 157 2.33 −7.12 20 160.23 2.31 −7.16 20 160.5 2.31 −7.16 20 160.97 2.30 −7.17

0.55 2.5 126.19 2.51 −8.76 0.55 2.5 126.17 2.51 −8.76 0.55 2.5 126.33 2.50 −8.76 0.55 2.5 128.2 2.49 −8.79

5 135.42 2.45 −8.21 5 136.53 2.44 −8.22 5 135.16 2.45 −8.20 5 138.69 2.43 −8.26

10 146.25 2.39 −7.67 10 148.19 2.37 −7.69 10 148.01 2.38 −7.69 10 148.62 2.37 −7.70

20 158.68 2.32 −7.14 20 162.23 2.30 −7.18 20 162.47 2.30 −7.19 20 162.53 2.30 −7.19

0.6 2.5 127.43 2.50 −8.78 0.6 2.5 127.4 2.50 −8.78 0.6 2.5 127.82 2.49 −8.78 0.6 2.5 129.68 2.48 −8.81

5 136.5 2.44 −8.22 5 138.1 2.43 −8.25 5 136.96 2.44 −8.23 5 140.3 2.42 −8.28

10 148.06 2.37 −7.69 10 150.02 2.36 −7.72 10 150.69 2.36 −7.73 10 150.11 2.36 −7.72

20 160.6 2.31 −7.16 20 164.32 2.29 −7.21 20 164.54 2.29 −7.21 20 164.9 2.28 −7.21

0.65 2.5 128.91 2.49 −8.80 0.65 2.5 128.89 2.49 −8.80 0.65 2.5 129.31 2.49 −8.81 0.65 2.5 131.43 2.47 −8.84

5 138.01 2.43 −8.25 5 139.97 2.42 −8.27 5 138.44 2.43 −8.25 5 142.13 2.41 −8.30

10 150.02 2.36 −7.72 10 151.98 2.35 −7.74 10 151.87 2.35 −7.74 10 151.92 2.35 −7.74

20 162.58 2.30 −7.19 20 166.53 2.28 −7.23 20 166.77 2.27 −7.24 20 166.87 2.27 −7.24

(Continues)

Polymer Composites, 2025

15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
 TABLE 2    |    (Continued)

0.0 wt. %F-­SWCNTs 0.05 wt. %F-­SWCNTs 0.1 wt. %F-­SWCNTs 0.2 wt. %F-­SWCNTs

α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2) α β Tα 1000/Tα ln(β/Tα 2)

0.7 2.5 130.4 2.48 −8.82 0.7 2.5 130.63 2.48 −8.83 0.7 2.5 130.79 2.48 −8.83 0.7 2.5 133.17 2.46 −8.87

5 139.84 2.42 −8.27 5 141.95 2.41 −8.30 5 140.4 2.42 −8.28 5 144.09 2.40 −8.33

10 151.98 2.35 −7.74 10 154.38 2.34 −7.78 10 154.6 2.34 −7.78 10 154.13 2.34 −7.77

20 164.9 2.28 −7.21 20 169.16 2.26 −7.27 20 169.25 2.26 −7.27 20 168.85 2.26 −7.26

0.75 2.5 131.9 2.47 −8.85 0.75 2.5 132.36 2.47 −8.85 0.75 2.5 132.77 2.46 −8.86 0.75 2.5 135.4 2.45 −8.90

5 141.31 2.41 −8.29 5 143.9 2.40 −8.33 5 142.37 2.41 −8.31 5 146.04 2.39 −8.36

10 153.94 2.34 −7.77 10 156.8 2.33 −7.81 10 156.55 2.33 −7.80 10 156.41 2.33 −7.80

20 167.71 2.27 −7.25 20 171.67 2.25 −7.30 20 172.33 2.25 −7.30 20 171.46 2.25 −7.29

0.8 2.5 133.88 2.46 −8.88 0.8 2.5 134.58 2.45 −8.89 0.8 2.5 134.75 2.45 −8.89 0.8 2.5 137.87 2.43 −8.94

5 143.3 2.40 −8.32 5 146.35 2.38 −8.36 5 144.81 2.39 −8.34 5 148.97 2.37 −8.40

10 156.88 2.33 −7.81 10 159.74 2.31 −7.84 10 159.49 2.31 −7.84 10 158.77 2.32 −7.83

20 170 2.26 −7.28 20 175.12 2.23 −7.34 20 175.45 2.23 −7.34 20 174.8 2.23 −7.33

0.85 2.5 136.36 2.44 −8.91 0.85 2.5 137.06 2.44 −8.92 0.85 2.5 137.47 2.44 −8.93 0.85 2.5 141.31 2.41 −8.99

5 146.22 2.39 −8.36 5 149.28 2.37 −8.40 5 147.25 2.38 −8.37 5 151.91 2.35 −8.44

10 159.83 2.31 −7.85 10 163.69 2.29 −7.89 10 163.42 2.29 −7.89 10 162.3 2.30 −7.88

20 173.74 2.24 −7.32 20 179.12 2.21 −7.38 20 180.3 2.21 −7.39 20 178.79 2.21 −7.38

0.9 2.5 140 2.42 −8.97 0.9 2.5 140.77 2.42 −8.98 0.9 2.5 141.16 2.41 −8.98 0.9 2.5 146.75 2.38 −9.06

5 149.15 2.37 −8.40 5 153.68 2.34 −8.46 5 151.17 2.36 −8.43 5 157.31 2.32 −8.51

10 164.78 2.28 −7.91 10 168.66 2.26 −7.95 10 168.37 2.27 −7.95 10 167.25 2.27 −7.94

20 178 2.22 −7.37 20 185.12 2.18 −7.45 20 186.3 2.18 −7.46 20 184.78 2.18 −7.44

0.95 2.5 146.47 2.38 −9.06 0.95 2.5 147.42 2.38 −9.07 0.95 2.5 147.54 2.38 −9.07 0.95 2.5 157.08 2.33 −9.20

5 154.54 2.34 −8.47 5 160.57 2.31 −8.55 5 156.55 2.33 −8.50 5 166.72 2.27 −8.62

10 171.76 2.25 −7.99 10 176.66 2.22 −8.05 10 177.36 2.22 −8.05 10 175.23 2.23 −8.03

20 185.71 2.18 −7.45 20 193.15 2.15 −7.53 20 196.32 2.13 −7.56 20 194.81 2.14 −7.55

7 of 16
15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
activation energy values based on different α levels, allowing for

ln(β/Tα2)
a more detailed and accurate analysis.

−8.72
−8.17

−7.09
−7.62
Current study reports the implementation of the GMN ap-
proach, which provides the use of isothermal DSC data to opti-

0.2 wt. %F-­SWCNTs

mize the pre-­exponential and activation energy factor. It focuses

1000/Tα
on a widely known methodology of modeling curing behavior,

2.46

2.34
2.51

2.4
but the introduction and implementation of the new algorithm,
which in turn has shown good agreement with the highly cited
Kissinger method. In this approach, experimental differential
scanning calorimetry (DSC) results for single-­wall carbon nano-

125

158
133
143
Tα
tubes impregnated epoxy resin were utilized. The GMN algo-
rithm's results were compared with other well-­known methods.

2.5

20
10
β

5
2   |   Experimental

2)
ln(β/Tα

−8.19
−8.74

−7.65
−7.13
The SWCNTs have an outer mean diameter of less than 2 nm and
a length exceeding 1 μm. They are of high quality, with a G/D
ratio greater than 90 and an iron content below 15%. A solvent-­

0.1 wt. %F-­SWCNTs

type epoxy, F-­R ES21 (Fibermak Kompozit in Izmir, Turkey) with
a viscosity of 400–650 mPa.s at 25°C was used as the resin sys-

1000/Tα

2.46

2.32
2.51
tem. Additionally, 800 tex carbon fiber (A-­49 filaments), supplied

2.4
TABLE 3    |    Results of Kissinger method for different carbon/epoxy prepreg systems. Units: β (°C/min), Tα (°C), 1000/Tα (1/K).
by DowAksa Inc., Turkey, was used as the reinforcing material.

F-­S WCNTs were produced by oxidizing the surface of SWCNTs

134
125

158
145
Tα
with carboxyl (-­COOH) groups through an acid treatment pro-
cess. The epoxy resin systems were modified with 0.05, 0.1, and
0.2 wt. % functionalized SWCNTs (F-­S WCNTs). These prepared

2.5

20
10
β

5
F-­S WCNT/epoxy nano composite blends were impregnated on
carbon fibers (CFs) using the drum-­t ype winding technique in
order to prepare prepreg systems.
2)
ln(β/Tα

−8.19
−8.72

−7.65
−7.13
Sample preparation was carried out at room temperature. Defects
were introduced on the sidewalls of single-­wall carbon nano-
0.05 wt. %F-­SWCNTs

tubes (SWCNTs) through an oxidative process using a mixture

of strong acids (HNO3 and H2SO4 in a 1:3 ratio), sourced from
1000/Tα

Sigma Aldrich, InterLab, Turkey. FTIR spectroscopy confirmed

2.46
2.52

2.32
2.4
the formation of carboxylic acid (–COOH) groups on the SWCNT
sidewalls as a result of this oxidation process. After undergoing
covalent functionalization, these SWCNTs were designated as F-­
SWCNTs. A multi-­step dispersion technique was then used to in-
124
134

158
145
Tα

tegrate the F-­SWCNTs into the epoxy resin system. Initially, the
F-­SWCNTs were sonicated in the solvent of the resin system (F-­
Prepreg) for 15 min. This solution was subsequently mixed into
2.5

20
10
β

the epoxy resin (RES-­21) and subjected to ultrasonication for

90 min. Following the addition of the hardener, the mixture was
2)

mechanically stirred at 2000 rpm for 15 min. The final resin sys-
ln(β/Tα
−8.72
−8.17

−7.09
−7.62

tem comprised epoxy (RES 21), a hardener (21 wt.% of the epoxy),
a solvent (10 wt.% of the total resin system), and F-­SWCNTs. This
modified resin system containing F-­SWCNTs was then prepared
for the fabrication of nano-­engineered prepregs. The sample
1000/Tα

preparation steps are depicted in Figure 1.

2.46

2.34
2.52

2.4
0.0 wt. %F-­SWCNTs

Laboratory-­scale unidirectional (UD) prepregs with and with-

out F-­S WCNTs were produced by a drum-­t ype winding process
(Figure 2) at room temperature. This process utilized a solution
124
133

155
143
Tα

impregnation of modified F-­S WCNTs in epoxy applied to car-

bon filaments. Fibermak Kompozit's winding machine enabled
the desired pattern to be applied to the mandrel as it was impreg-
2.5

20
10
β

nated in a bath containing the modified or unmodified resin.

8 of 16 Polymer Composites, 2025

 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The hoop winding pattern, with a winding angle of 90°, resulted d𝛼 (dH∕dt)t
in prepreg products. The resing at 750–850 g was employed for = (2)
dt Δ HR
production. Resin reached to B-­stage in the drying oven by sol-
vent evaporation after winding prepreg. where Ht is the amount of heat released in time t and (dH∕dt)t is
the rate of the heat generation and Δ HR is the total heat of reac-
Temperature variations during the curing process play a crucial tion. It is commonly accepted in the kinetic analysis of chemi-
role in determining the degree of cure for the prepreg system, cal reactions by thermal analysis that any chemical process of
which is closely linked to the heat of reaction. These parame- reaction obeys a rate law of the form given in Equation (3) as:
ters are measured using differential scanning calorimetry (DSC)
and are essential for assessing the quality of the final material. d𝛼
= k(T) f (𝛼) (3)
For this analysis, 3–6 mg samples of the mixtures were placed in dt
aluminum pans, with a nitrogen purge rate of 50 mL/min using
a DSC 50 Series from Schimadzu Inc., Japan. Dynamic scans Here, dα/dt represents the rate of reaction, k(T) is the
were conducted at various heating rates (2.5, 5, 10, and 20°C/ temperature-­dependent rate constant, and f(α) corresponds to
min) from room temperature up to 250°C in a nitrogen atmo- the reaction model [28]. The temperature dependence of the re-
sphere. The upward peaks observed in the DSC curves indicate action rate is typically described by an Arrhenius expression [14]:
that the reaction was exothermic. ( )
E
k(T) = A exp − (4)
RT
The curing reaction of an epoxy resin is an exothermic process.
In DSC measurements, the degree of cure, denoted as α, ranges
In this expression, A represents the pre-­exponential factor, E is
from 0, representing a completely uncured state, to 1, indicating
the activation energy, R refers to the universal gas constant, and
a fully cured state. The degree of cure is mathematically defined
T corresponds to the absolute temperature. This relationship is
in Equations (1) and (2) as follows:
expressed in Equation (4). In an isothermal process, the tempera-
Ht ture remains constant, whereas in a non-­isothermal process, the
𝛼= (1) temperature typically increases at a constant heating rate.
ΔH

FIGURE 5    |    Different carbon/epoxy prepreg systems at different F-­S WCNTs weight ratios of (a) 0.0, (b) 0.05, (c) 0.1, and (d) 0.2 based on KAS
method.

9 of 16
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
dT 3   |   A New Numerical Approach: Nonlinear Fitting
𝛽=
dt (5)
Suppose that the experimental data for temperature, T = [T1, T2
where β is the heating rate. Combining Equations (3), (4), and (5) …, TN] and conversion values, α = [y1, y2, …, yN] are given where
yields a resulting equation for non-­isothermal conditions: N is the number of the data. In this study, the following model
( ) given in Equation (9) is considered for the activation function:
d𝛼 A E
= exp − f (𝛼) (6)
dT 𝛽 RT 𝛼(T) = A(tanh(B(T − C) − 3) + 1) (9)

Hence, if the reaction model that describes the data is where A, B, and C are unknown. A function that admits a form
known, the conversion at a specific temperature or heating given in Equation (9) was sought to fit the best given data. The
rate can be predicted using the equations provided above. motivation for choosing the hyperbolic tangent function is that
The activation energy for the curing reaction can be deter- the graph of the experimental conversion values is similar to this
mined using iso-­conversional methods, such as the Kissinger-­ function. To minimize the difference between the experiments
Akahira-­Sunose (KAS) and Kissinger methods, which and the function of activation, we employed the nonlinear least
are among the most frequently cited in the literature. The squares method. The requirement of the nonlinear least squares
Kissinger-­A kahira-­Sunose (KAS) method is represented by procedure is to find a minimum of the error function given in
Equation (7) as follows: Equation (10).
( ) ∑N ( ( ( ( ) ) ))2
𝛽i E𝛼 E(A, B, C) = yn − A tanh B Tn − C − 3 + 1 (10)
ln 2
= Const − (7) n=1
T𝛼,i R T𝛼
The partial derivative E(A, B, C) with respect to A, B, C must be
and the Kissinger Method is given in Equation (8) as: zero in order to guarantee the minimum of the function given in
Equation (10).
( )
𝛽i E
ln = Const − (8) 𝜕E 𝜕E 𝜕E
Tp2 R Tp = 0, = 0, = 0. (11)
𝜕A 𝜕B 𝜕C

FIGURE 6    |    Different carbon/epoxy prepreg systems at different F-­S WCNTs weight ratios of (a) 0.0, (b) 0.05, (c) 0.1, and (d) 0.2 based on Kissinger
method.

10 of 16 Polymer Composites, 2025

 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
� �
When all partial derivatives in Equation (11) are set equal to ⎛F z k ⎞
zero, the resultant normal equations are obtained: � �⎜ � �⎟
z k+1 = z k − J −1 z k ⎜G z k ⎟ (15)
⎜ � k�⎟
𝜕E ∑ (
N
( ( ) ) ) ⎝S z ⎠
= −2 tanh B Tn − C − 3 + 1
𝜕A n=1 (12)
( ( ( ( ) ) )) ⎛ Ak ⎞
yn − A tan h B Tn − C − 3 + 1 ⎜ ⎟
z k = ⎜ Bk ⎟ (16)
⎜ k⎟
⎝C ⎠
𝜕E ∑
N
( ( ( ) ))2 ( )
= −2A sech B Tn − C − 3 Tn − C And J −1 the inverse of the Jacobian matrix is in Equation (17) as,
𝜕B n=1 (13)
( ( ( ( ) ) ))
yn − A tan h B Tn − C − 3 + 1 ⎛ 𝜕E 𝜕E 𝜕E ⎞
J(z) = ⎜ 𝜕A 𝜕B 𝜕C ⎟. (17)
⎜ ⎟
⎝ ⋮ ⋱ ⋮ ⎠
𝜕E ∑
N
( ( ( ) ))2
= 2AB sech B Tn − C − 3
𝜕C n=1 (14) To start Newton's iterative method, an initial condition is needed.
( ( ( ( ) ) )) For convergent A, B, C values, an appropriate initial condition is
yn − A tan h B Tn − C − 3 + 1
chosen in Equation (18) as,

The equations in Equations (12–14) are the nonlinear system of ⎛ 0.5 ⎞

equations with respect to unknowns A, B, and C. The resultant ⎜ ⎟
0
z = ⎜0.05⎟ (18)
system can be solved by Newton's method m which is given in ⎜ ⎟
Equations (15) and (16). ⎝ 380 ⎠

TABLE 4    |    Curing results of systems at different % F-­S WCNTs ratios.

Eα (kJ/mol)
Neat 0.05 wt. %F-­SWCNTs 0.1 wt. %F-­SWCNTs 0.2 wt. %F-­SWCNTs
Kissinger method 73.6 66.3 68.1 70.8
KAS method α (%)
0.1 75.5 69.0 56.6 63.6
0.15 75.5 68.0 61.1 64.9
0.2 74.6 66.7 61.5 66.2
0.25 74.8 66.3 62.2 68.1
0.3 73.2 65.9 62.4 67.9
0.35 73.8 65.4 62.7 68.9
0.4 73.0 64.7 62.7 68.9
0.45 72.2 64.4 62.5 68.9
0.5 71.1 63.7 62.3 68.4
0.55 71.3 63.2 61.9 68.8
0.6 69.7 62.0 61.0 68.5
0.65 68.9 61.3 60.4 67.8
0.7 67.8 60.3 59.3 68.1
0.75 65.4 59.5 58.0 68.0
0.8 65.0 58.1 56.8 67.5
0.85 63.8 56.3 53.8 67.4
0.9 62.2 54.2 51.6 68.6
0.95 61.5 54.3 47.4 73.0

11 of 16
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The Newton method is generally a convergent scheme if the given in Equation (9) fits the α(T). X. Ramis et al. [29] reported a
initial condition is chosen appropriately. The convergence of various choice of f(α). For the reaction model, the most suitable
the method was also checked by using the tolerance in our al- choice was found as.
gorithm. The process continued until the difference of the two
consecutive iterations is less than the tolerance. This guarantees f (𝛼) = 𝛼 m (1 − α)n (19)
that the sequences defined by Newton's Method are a Cauchy
sequence. Therefore, it is convergent. where m + n = 2. So, in this study, n = 1.4, m = 0.6 (m + n = 2) were
found to be the best choice for the reaction model. Several combi-
nations of m and n values were also tried so that the experimental
3.1   |   Data Linearization Technique data fit the numerical solution of the ODE after finding the param-
eters A and E. Then, the parameters m and n were optimized by
In this section, values of pre-­experimental factor A and activa- this approach. As a result, there are only two unknowns A and E
tion energy E will be sought which are given in Equation (6). in Equation (6). To find these unknowns, the data linearization
There are four unknowns, and they are A, E, α and f(α) in technique was used. The first step is to take the logarithm of both
Equation (6). By using the curve fitting procedure, the function sides of Equation (6) as given in Equation (20).

FIGURE 7    |    Variation of E α estimated with (a) KAS and (b) Kissinger methods.

12 of 16 Polymer Composites, 2025

 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
( ) ( )
𝛽 dα E
ln + ln = ln(A) − . (20)
f(α) dT RT

This results in a linear relation between the new values X and Y

given in Equation (21):

Y = A + EX. (21)

Then the change of variables given in Equation (22) is

introduced.
( ) ( )
𝛽 d𝛼 1
Y = ln + ln , A = ln(A), X =− (22)
f (α) dT RT

GMN Algorithm is briefly summarized as follows.

• Step 1: Approximate α(T) as a closed function given in

FIGURE 8    |    The α(T) as a function of temperature estimated using Equation (9) by using the experimental data.
the activation function with fitting compared to experiment.

TABLE 5    |    The parameters for different heating rates.

Sample ID with CNT content Sample no β A B C E ln(A)

0.00 wt.% 1 2.5 0.5 0.0765 358.3126 58.712 12.6856
0.00 wt.% 1 5 0.5 0.0767 367.8309 63.545 14.4190
0.00 wt.% 1 10 0.5 0.0608 368.3936 57.664 12.6843
0.00 wt.% 1 20 0.5 0.0564 377.0870 55.267 12.1549
0.05 wt.% 2 2.5 0.5 0.0732 356.5693 53.8536 10.6157
0.05 wt.% 2 5 0.5 0.0619 359.3702 53.8536 11.3089
0.05 wt.% 2 10 0.5 0.0558 365.7544 52.9976 11.2936
0.05 wt.% 2 20 0.5 0.0509 374.5726 52.9925 11.2847
0.1 wt.% 3 2.5 0.5 0.0669 352.6327 59.6918 12.8780
0.1 wt.% 3 5 0.5 0.0627 358.7051 59.6918 12.9615
0.1 wt.% 3 10 0.5 0.0545 364.2501 53.7177 11.3744
0.1 wt.% 3 20 0.5 0.0515 375.6366 54.5422 11.6969
0.2 wt.% 4 2.5 0.5 0.0651 353.4174 51.0417 10.0880
0.2 wt.% 4 5 0.5 0.0616 361.2695 56.0027 10.0880
0.2 wt.% 4 10 0.5 0.0596 369.6381 59.9492 13.2245
0.2 wt.% 4 20 0.5 0.0561 380.6521 54.9921 11.8906

TABLE 6    |    The results of the GMN method compared to current well-­k nown techniques.

Kissinger method KAS method Optimized GMN method

Sample ID with CNT content E (kJ/mol) ln(A) E (kJ/mol) ln(A) E(kJ/mol) ln(A)
0.00 wt.% 73.6 13.6 75.5–61.5 15.2–10.38 63.515–55.267 14.4190–12.1549
0.05 wt.% 66.3 11.3 69.0–54.3 13.3–8.2 52.9925–53.8536 10.6157–11.2847
0.1 wt.% 68.1 11.8 61.1–51.6 10.6–7.8 53.7177–59.6918 11.6969–12.9615
0.2 wt.% 70.8 12.6 64.9–68.6 11.6–10.6 51.0417–59.9492 10.0880–11.7959

13 of 16
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
• Step 2: Use nonlinear least square method to set up alge- thermal diffusivity also improves, owing to the exceptional ther-
braic equations in terms of constants A, B, and C. mal conductivity of the nanotubes.
• Step 3: Use the Newton iterative method to find A, B, and C
The crosslinking reaction is increased due to modified nanotubes
in the nonlinear algebraic equations obtained in step 2.
with a high concentration of OH surface groups at 0.05% and 0.1%
• Step 4: Insert closed form of the α(T) and f(α) = αm (1−α)n weight ratio cases. Increasing nanotube addition created higher
in Equation (6) where m and n values m = 1.6, n = 0.4 viscosity and agglomeration, which yielded lower cure enthalpy.
(m + n = 2) into the equations. The dispersion homogeneity is key for the formation of a perco-
lated network more easily than in agglomerated systems.
• Step 5: Take the log of both sides of the Equation (6) in order
to use linear fitting for parameters A and E.
Figure 7 illustrates the relationship between the activation en-
• Step 6: Optimize the parameters A and E by solving the ODE ergy (E α) and the degree of conversion (α) during the curing
given in Equation (6) until the error between the experimen- process for various prepreg systems. The energy for activation
tal data and numerical solution of the ODE is minimum. stayed constant. There is a good agreement between the E α val-
ues determined by the isoconversional method (KAS) and the
Kissinger method, despite inherent experimental errors. The
4   |   Result and Discussion relatively lower Eα values calculated by the KAS method could
be due to its better approximation for the temperature integral.
4.1   |   Isoconversional Methods

The total heat of reaction for the CF/epoxy prepreg systems, both 4.2   |   The Optimized Parameter Approach
with and without F-­SWCNTs, was directly measured using DSC.
Figure 3 presents representative DSC curves for these systems at a In this study, we conducted kinetic analysis using an
heating rate of 2.5°C/min. The values for the total heat of reaction, Optimized Parameter Approach that combines nonlinear least
with and without F-­SWCNTs, are summarized in Table 1. This squares fitting of a tanh function with a linearization tech-
heat, which correlates with the extent of curing, indicates that the nique. This novel algorithm was validated using experimental
prepreg with the modified resin system containing F-­SWCNTs DSC results. GMN algorithm was then compared with those
achieved a higher degree of cure. Figure 4 shows the relation- from other methods documented in the literature. First, the
ship between conversion and temperature for various heating conversion function (α(T)) was described as Equation (9). To
rates. The conversion degrees were determined by integrating start Newton's iterative method, an initial condition is needed.
the exothermic calorimetric peaks, as described by Equation (1). For convergent A, B, C values, an appropriate initial condi-
The plots display a characteristic sigmoidal shape, with the con- tion was chosen. For calculations, the Matlab solver was used.
version (α) rising slowly at both the start and end of the reaction
and increasing rapidly during the intermediate phase. The curing
TABLE 7    |    The error between the numerical solution of ODE after
process begins at a slow rate as the resin rapidly nears its gelation
optimization of the parameters A and E and experimental DSC data.
temperature. The rate of reaction increases when the gel resin is
formed. The network structure's crosslinking advances during the Sample ID with CNT content β l∞
intermediate stage, slowly converting the resin into a glassy state.
0.00 wt.% 2.5 0.0214
As crosslinking intensifies, the curing rate diminishes. In the final
phase, as the degree of cure approaches its maximum, the curing 0.00 wt.% 5 0.0142
rate further declines due to the system's increased viscosity.
0.00 wt.% 10 0.0076
The curing process at different heating rates is studied to de- 0.00 wt.% 20 0.0105
termine the conversion-­dependent activation. For specific con-
0.05 wt.% 2.5 0.0134
version values, the corresponding reciprocal temperatures are
identified at heating rates of 2.5, 10, and 20°C/min and plotted 0.05 wt.% 5 0.0134
according to Equations (7–8). This analysis evaluates and mod- 0.05 wt.% 10 0.0041
els the reaction kinetics within the conversion range of 0.1 to
0.95, with intervals of 0.05. The cure kinetic parameters were 0.05 wt.% 20 0.0030
extracted using OriginPro software through the linear least 0.1 wt.% 2.5 0.0158
squares method. The results obtained from these algorithms are
detailed in Tables 2 and 3. The slope of the straight lines depicted 0.1 wt.% 5 0.0166
in Figures 5 and 6 is used to determine the activation energy for 0.1 wt.% 10 0.0058
the corresponding prepreg system.
0.1 wt.% 20 0.0109
Table 4 presents the average values of activation energy (Eα) 0.2 wt.% 2.5 0.0157
derived from both the original Kissinger method and the iso-­
conversional KAS method. Lower activation energies are ob- 0.2 wt.% 5 0.0135
served compared to the pure epoxy resin prepreg. Curing 0.2 wt.% 10 0.0057
efficiency increased due to increased thermal conductivity by
0.2 wt.% 20 0.0104
F-­S WCNTs addition. As the weight fraction of CNTs increases,

14 of 16 Polymer Composites, 2025

 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
scanning calorimetry. Kinetic analysis was performed by iso-­
conversional methods, and an Optimized Parameter Approach
combining a tanh nonlinear least square fitting and linearization
named the GMN method. So, this study reports the implementa-
tion of the GMN approach taking use of isothermal DSC data to
optimize the pre-­exponential and activation energy factor. For
all investigated hybrid composites, the activation energy was
found to be lower than neat epoxy resin prepreg (73 kJ/mol). The
energy for activation is reduced due to increased thermal con-
ductivity by CNT addition. In the experimental study, function-
alized single-­wall carbon nanotubes oxidized by carboxyl groups
were obtained. The enhanced dispersion of the nano particulates
was presented using an acid treatment. Carbon fibers (CFs) were
impregnated with modified F-­SWCNTs at 0.05, 0.1, and 0.2wt.%.
Experimentally determined DSC data sets for CF/epoxy prepregs
containing F-­SWCNTs were employed. The GMN algorithm's re-
sults obtained compare to KAS and Kissinger methods. Results
of GMN agreed well with the Kissinger method, which is the
FIGURE 9    |    Temperature versus conversion (α) values when most cited algorithm in the literature. Compared to the results
lnA = 12.6843 and E = 57.664. of the KAS method, which changes depending on α, the mean
values of the GMN were found to be within the given range. As a
result, the GMN method was revealed as an accurate algorithm
Finally, the unknown A, B, and C parameters were obtained
to obtain the activation energy.
as 0.5, 0.0765, and 358.3126, respectively. The α(T) values ob-
tained experimentally were compared with the least square
approximation as a function of temperature in Figure 8. The
findings presented in this figure may also be considered as Acknowledgments
evidence of the convergence of the nonlinear system given in This work was supported by the Scientific and Technological Research
Equations (12–14). Council of Turkey (TUBITAK) under Grant Number 218M701. The au-
thors acknowledge IZTECH Center for Materials Research (MAM) for
Next, values of pre-­experimental factor A and activation en- providing testing services in this study.
ergy E given in Equation (6) are determined. There are four
unknowns: A, E, α and f (α). The function of α(T) is obtained, Author Contributions
and f (α) = αm(1 − α)n, n = 1.4, m = 0.6 (m + n = 2) is the best The authors takes full responsibility for this article.
choice for the reaction model. In Table 5, A, B, C values to-
gether with E and ln(A) values are given for various carbon/
epoxy prepreg systems for different heating rates. The average Conflicts of Interest
of activation energy for the whole process can be estimated The authors declare no conflicts of interest.
using the Kissinger method, including α values. On the other
hand, the KAS method changes depending on α as it is an iso-­ Data Availability Statement
conversional method. The comparison of the optimum values
The data that support the findings of this study are available on request
of the Optimized GMN method with Kissinger and KAS meth-
from the corresponding author. The data are not publicly available due
ods is given in Table 6. In summary, the optimized GMN algo- to privacy or ethical restrictions.
rithm results agreed with experiments.

In addition, the error between the numerical solution of ODE References

after optimization of the parameters A and E and experimental 1. F. Buehler and J. Seferis, “Effect of Reinforcement and Solvent Con-
DSC data was calculated for various samples as given in Table 7. tent on Moisture Absorption in Epoxy Composite Materials,” Compos-
So, the results revealed the accuracy of the parameters A and E, ites Part A: Applied Science and Manufacturing 31, no. 7 (2000): 741–748.
up to a certain degree. 2. N. A. Siddiqui, S. U. Khan, P. C. Ma, C. Y. Li, and J.-­K . Kim, “Man-
ufacturing and Characterization of Carbon Fibre/Epoxy Composite
As an example, Figure 9 shows the conversion (α) values by Prepregs Containing Carbon Nanotubes,” Composites Part A: Applied
temperature when lnA = 12.6843 and E = 57.664. The findings Science and Manufacturing 42, no. 10 (2011): 1412–1420.
prove that the experimental and numerical values are in very 3. L. Xu, J. Fu, and J. Schlup, “In Situ Near-­Infrared Spectroscopic In-
good agreement. vestigation of Epoxy Resin-­Aromatic Amine Cure Mechanisms,” Jour-
nal of the American Chemical Society 116, no. 7 (1994): 2821–2826.
4. K. Kumar and J. P. Davim, eds., Hierarchical Composite Materials:
5   |   Conclusion Materials, Manufacturing, Engineering (De Gruyter, 2018).
5. V. Bavastrello, Z. Jiang, C. Nicolini, J. Schjødt-­T homsen, S. K. Srivas-
The non-­isothermal curing of carbon-­fiber/epoxy prepreg sys- tava, and D. K. Tripathy, Nanocomposites: Materials, Manufacturing
tems with/without F-­SWCNTs was investigated by differential and Engineering (Walter de Gruyter, 2013).

15 of 16
 15480569, 0, Downloaded from https://4spepublications.onlinelibrary.wiley.com/doi/10.1002/pc.30039 by Bangladesh Hinari NPL, Wiley Online Library on [10/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6. P.-­C. Ma, N. A. Siddiqui, G. Marom, and J.-­K . Kim, “Dispersion and Analysis of Thermal Polymerization Kinetics,” Thermochimica Acta
Functionalization of Carbon Nanotubes for Polymer-­Based Nanocom- 714 (2022): 179243.
posites: A Review,” Composites Part A: Applied Science and Manufactur-
25. S. Heinze and A. T. Echtermeyer, “A Practical Approach for Data
ing 41, no. 10 (2010): 1345–1367.
Gathering for Polymer Cure Simulations,” Applied Sciences 8, no. 11
7. M. H. Alma, Z. Candan, O. Gonultas, et al., Green Composites: Mate- (2018): 2227.
rials, Manufacturing and Engineering (Walter de Gruyter GmbH & Co
26. A. Hosseinpour, H. Nazockdast, T. Behzad, and H. SalimiJazi, “In-
KG, 2017).
vestigation of the Cure Kinetics of an Epoxy Resin by Advanced Isocon-
8. J. P. Davim, ed., Statistical and Computational Techniques in Manu- versional and Model-­Fitting Methods,” in AIP Conference Proceedings,
facturing (Springer Science & Business Media, 2012). vol. 1713 (AIP Publishing LLC, 2016), 110004.
9. J. P. Davim and J. P. Davim, eds., Computational Methods and Pro- 27. M. Oz, “A Numerical Approach for Optimization of Curing Kinetics
duction Engineering: Research and Development (Woodhead Publishing, of Composite Material,” 2021 MS Thesis, Izmir Institute of Technology.
2017).
28. P. C. Ma, J.-­K . Kim, and B. Z. Tang, “Effects of Silane Functional-
10. H. E. Kissinger, “Reaction Kinetics in Differential Thermal Analy- ization on the Properties of Carbon Nanotube/Epoxy Nanocomposites,”
sis,” Analytical Chemistry 29, no. 11 (1957): 1702–1706. Composites Science and Technology 67, no. 14 (2007): 2965–2972.
11. M. Yücel, “Fabrication of Thin Layer Polymer-­ Based Biointer- 29. X. Ramis, J. M. Salla, A. Cadenato, and J. M. Morancho, “Simulation
phase for Biosensing Application,” 2016 MS Thesis, I˙zmir Institute of of Isothermal Cure of a Powder Coating Non Isothermal DSC Experi-
Technology. ment,” Journal of Thermal Analysis and Calorimetry 72 (2003): 707–718.
12. S. W. Lee, B.-­S. Kim, S. Chen, Y. Shao-­Horn, and P. T. Hammond,
“Layer-­By-­Layer Assembly of All Carbon Nanotube Ultrathin Films for
Electrochemical Applications,” Journal of the American Chemical Soci-
ety 131, no. 2 (2009): 671–679.
13. T. Hsieh, A. Kinloch, K. Masania, J. S. Lee, A. Taylor, and S.
Sprenger, “The Toughness of Epoxy Polymers and Fibre Composites
Modified With Rubber Microparticles and Silica Nanoparticles,” Jour-
nal of Materials Science 45, no. 5 (2010): 1193–1210.
14. W.-­J. Chen, Y.-­L . Li, C.-­L . Chiang, et al., “Preparation and Charac-
terization of Carbon Nanotubes/Epoxy Resin Nano-­Prepreg for Nano-
composites,” Journal of Physics and Chemistry of Solids 71, no. 4 (2010):
431–435.
15. S. Reich, C. Thomsen, and J. Maultzsch, Carbon Nanotubes: Basic
Concepts and Physical Properties (John Wiley & Sons, 2008).
16. H. A. Khonakdar, V. Goodarzi, D. Puglia, et al., “Calorimetric Anal-
ysis and Molecular Dynamics Simulation of Cure Kinetics of Epoxy/
Chitosan-­Modified fe3o4 Nanocomposites,” Progress in Organic Coat-
ings 112 (2017): 176–186, https://​doi.​org/​10.​1016/j.​porgc​oat.​2 017.​07.​015.
17. S. H. Ryu, J. Sin, and A. Shanmugharaj, “Study on the Effect of
Hexamethylene Diamine Function-­Alized Graphene Oxide on the Cur-
ing Kinetics of Epoxy Nanocomposites,” European Polymer Journal 52
(2014): 88–97.
18. T. Ozawa, “Kinetic Analysis of Derivative Curves in Thermal Analy-
sis,” Journal of Thermal Analysis 2, no. 3 (1970): 301–324.
19. S. Bal and S. Samal, “Carbon Nanotube Reinforced Polymer Com-
posites—A State of the Art,” Bulletin of Materials Science 30, no. 4
(2007): 379.
20. V. Slov'ak and B. Taraba, “Urea and cacl2 as Inhibitors of Coal Low-­
Temperature Oxidation,” Journal of Thermal Analysis and Calorimetry
110, no. 1 (2012): 363–367.
21. T. Dubaj, Z. Cibulkov'a, and P. Šimon, “An Incremental Isoconver-
sional Method for Kinetic Analysis Based on the Orthogonal Distance
Regression,” Journal of Computational Chemistry 36, no. 6 (2015):
392–398.
22. D. Trache, A. Abdelaziz, and B. Siouani, “A Simple and Linear Iso-
conversional Method to Determine the Pre-­E xponential Factors and
the Mathematical Reaction Mechanism Functions,” Journal of Thermal
Analysis and Calorimetry 128, no. 1 (2017): 335–348.
23. M. Stanko and M. Stommel, “Kinetic Prediction of Fast Curing Poly-
urethane Resins by Model-­Free Isoconversional Methods,” Polymers 10,
no. 7 (2018): 698.
24. S. Vyazovkin, D. Achilias, X. Fernandez-­Francos, A. Galukhin, and
N. Sbirrazzuoli, “ICTAC Kinetics Committee Recommendations for

16 of 16 Polymer Composites, 2025