A consistent cure kinetic model for AS4/3502 graphite/epoxy
D.D. Shin, H.T. Hahn
Mechanical and Aerospace Engineering Department, University of California, Los Angeles, CA 90095-1597, USA
Accepted 12 January 2000
Accurate cure kinetic model is crucial for correctly identifying the amount of heat generated for composite process simulation. This paper
presents a new technique for identifying cure kinetics parameters for Hercules AS4/3502 prepreg by normalizing the DSC data. The cure
kinetics is based on an autocatalytic model for the proposed method, which uses dynamic and isothermal DSC data to determine its
parameters. Existing models are also used to determine kinetic parameters but rendered inadequate because of the material’s temperature
dependent ﬁnal degree of cure. The model predictions determined from the new technique showed good agreement to both isothermal and
dynamic DSC data. The ﬁnal degree of cure was also in good agreement with experimental data. ᭧2000 Elsevier Science Ltd. All rights
Keyword: Cure kinetic
Cure kinetic model is an integral part of composite
process simulation, which is used to predict degree of cure
and amount of heat generation. The reaction order (nth
order) and autocatalytic (Sestak–Berggren) models are
most frequently cited among the existing models. The para-
meters involved in those models are usually determined
empirically from isothermal or dynamic differential
scanning calorimetry (DSC) data. An example of such a
method is observed for Hercules 3501-6 epoxy resin
where least-squares analysis is used to determine pre-
exponential factors, reaction orders, and activation energies
from isothermal DSC data .
As for Fiberite 977-2 resin, the nth order reaction model
was used to describe the reaction rate . The pertaining
kinetic parameters were determined from both isothermal
and dynamic DSC data using linear least-squares regression.
Good model-experiment correlation was observed when
model prediction was generated with parameters determined
at identical scanning conditions. Unfortunately, signiﬁcant
deviation was observed when dynamic model prediction
was generated with parameters extracted from isothermal
data. Same was true for isothermal model prediction
generated with kinetic parameters determined from
dynamic DSC data. Other material such as Fiberdux 6376
prepreg used modiﬁed nth order reaction model to accommo-
date temperature dependent ﬁnal degree of cure . The
kinetic parameters for this system were extracted from
isothermal DSC data using nonlinear least-squares analysis.
Although good model-experiment correlation was observed
for isothermal cases, the predictions were not in good agree-
ment with dynamic DSC data.
Malek stated that kinetic parameters could not be deter-
mined from a single DSC curve unless prior knowledge of
the kinetic model or at least one kinetic parameter was
known [4–6]. Similar viewpoints were shared by Montserrat
et al. who suggested that all kinetic parameters could not be
calculated from a single non-isothermal DSC curve by
conventional regression algorithms [7,8]. Fortunately, the
dilemma was resolved when the activation energy was
determined in prior via the Kissinger method .
Pollard and Kardos  determined kinetics model for
Hercules 3501-6 and 3502 resin systems using an Avrami
phase-change theory. The model was based on isothermal
DSC data and described the cure only up to the gel point.
Others studied the inﬂuence of reaction order and auto-
catalytic models at various stages of cure, and derived a
kinetic model that incorporated both reaction mechanisms
For the present paper, a cure kinetic model for Hercules
AS4/3502 prepreg was presented. Unfortunately, this
prepreg system had a temperature dependent ﬁnal degree
of cure that made the readily available kinetic models
incompatible. Simply applying temperature dependent
Composites: Part A 31 (2000) 991–999
1359-835X/00/$ - see front matter ᭧ 2000 Elsevier Science Ltd. All rights reserved.
* Tel.: ϩ 1-310-206-8157; fax: ϩ 1-310-825-2383.
E-mail address: email@example.com (H.T. Hahn).