Composite Materials

(Italy)

Part I

Composites with Organic Matrices

Modelling of Processing Technologies for Polymer Matrix Composites

Jose M. Kenny    Department of Materials and Production Engineering, University of Naples, P. Tecchio, 80125, Naples, ITALY

Abstract

A general model for the description of the thermo-chemo-rheological changes of the matrix during the processing of polymer based composites, has been developed in recent years. The master model includes, specific submodels dedicated to the different physico-chemical aspects of composite processing: reaction kinetics, heat transfer, viscosity, fluid flow and void formation during processing. The main objective of this work is to highlight the contribution of the understanding of the matrix behaviour to the right choice of processing conditions.

MODELLING PRINCIPLES FOR COMPOSITES PROCESSING

In the last decade several studies of the fundamental aspects of the processing of high performance composites have been reported [1–6]. In those research works a common scientific approach based on the understanding of the fundamental chemical and physical phenomena governing the behaviour of the composite matrix and allowing a better choice and control of processing conditions has been developed.

The final objective of those activities was the construction of a general processing model that could be adapted to different specific technologies. In order to develop such general model several submodels are needed as shown in Fig. 1. The first submodel describes the kinetics of the matrix chemical transformations, responsible of the final structure of the composite. The thermokinetic model predicts the exothermal heat of reaction and the degree of cure as a function of process time and temperature. The rheological model describes the viscosity evolution as function of time, temperature and degree of cure. Therefore, the rheological model is combined with the thermokinetic model forming the chemorheologicał model including also the gel point. For non-isothermal conditions also a heat transfer model is needed. If the heat transfer model is combined with the chemorheologicał and cure kinetic models, the degree of cure, temperature and viscosity, as a function of time and position in the composite, can be predicted. The flow model predicts resin content distribution and final composite thickness. Finally, the void model predicts the conditions needed to avoid the formation of voids.

Figure 1 Modeling philosophy of the processing of composite materials

A more detailed examination of the modelling of composite processing technologies will be given in the following sections where each of the submodels mentioned in this section, integrated into a general master model, will be discussed.

THERMOKINETIC MODEL

The thermokinetic model is the first step in the construction of the master model being a prerequisite for all the other submodels. It describes the rate at which heat is given off during the reaction, and the degree of cure, α, as a function of temperature and time. For curing reactions, the rate of heat flow dH/dt, can be conveniently measured by differential scanning calorimetry (DSC) [7]. It has been assumed for these processes that the quantity dH/dt is directly proportional to the rate of disappearance of reactive groups during curing, dα/dt. Empirical thermokinetic models for non-autocatalytic reactions have usually the following form:

   (1)

Integration of this equation can be used to predict α, dα/dt and dH/dt as a function of time and temperature. The rate constant K depends on temperature and generally is given by:

   (2)

KO is a preexponencial factor (frequency factor), E is the activation energy, R is me gas constant and T the absolute temperature. A much better description of the thermoset behavior has recently been reported [8]:

   (3)

where αm is the maximum degree of reaction obtained at a given temperature in an isothermal test. For reactions with autocatalytic behavior the following rate equation has been found useful [9]:

   (4)

where K1 and K2 are the kinetic constants depending on the temperature through an Arrhenius type equation like Eq. (2).

Although there are several ways to determine the parameters of the kinetic equation [7], one procedure is to perform isothermal DSC tests and then use nonlinear regression analysis to determine the constants [8]. The model can then be verified by comparison with results obtained in isothermal and in dynamic tests performed at different heating rates.

A practical application of the thermokinetic model is the prediction of the cure time at a given temperature to reach a specified degree of cure, for example the end of the cure or the gel point in combination with viscosity measurements. The thermokinetic model can be also used associated with the chemorheological model to describe the evolution of the polymeric structure and its viscosity. Finally, the thermokinetic model also predicts the amount of heat given off during the reaction. Then, if it is coupled with a heat transfer model (as described later), the temperature and degree of cure distribution inside a composite can be described for non-isothermal conditions.

RHEOLOGICAL MODEL

The rheological model should be able to predict the viscosity as a function of temperature and time. Since the viscosity depends on the degree of cure, the viscosity model has to be combined with the thermokinetic model, in the chemorheological model.

Viscosity measurements are usually performed in a dynamic mechanic spectrometer for fluids provided with disposable parallel plates. The location of the gel point can be determined from infinite viscosities obtained in shear mode tests [10]. Also it has been determined as the minimum in the loss factor (tan delta) at low frequencies [6] in dynamic tests.

There are two different phenomena which govern the viscosity of a thermoset [10]. One phenomenon is the growing size of the molecules during curing which increases the viscosity of the resin. The other is the effect of temperature on molecular mobility. These two mechanisms together determine the viscosity of reactive polymers in the fluid state. A fundamental approach has been reported [5,10]. The temperature effect can be taken into account using the WLF-equation while the molecular size dependence is considered a function of the molecular weight (Mw). Following this approach the final expression of the viscosity becomes:

   (5)

The parameter g is the ratio of the square of the radii of gyration of a branched to a linear chain of the same molecular weight; g is a number on the order of 1 that can be calculated applying statistical analysis. Since Tg depends on the degree of cure this dependence has to be determined and included into the expression. Equation (6) was found to successfully describe the experimentally determined viscosity of a model TGDDM-DDS epoxy system [10] and of TGDDM-DDS matrices of commercial prepregs [5]. The described theoretical approach, based on fundamental physical and chemical principles, requires the knowledge of the molecular weight and of the functionalities of the components, the nature of the chemical reactions and how Tg varies with the degree of cure.

For thermosets with complicated reaction mechanisms or where the composition and functionalities of the molecules in the resin mixture are unknown, an empirical approach is necessary. In order to describe the viscosity of polyester matrices Kenny et al. [6] adopted a model similar to the one originally used by Castro and Macosko for polyurethanes viscosity [11]:

   (6)

where μ is the viscosity, ag is the extent of reaction at the gel point and Aμ, Eμ, A and B are constants to be determined by regression analysis of experimental data.

In addition to the prediction of minimum viscosity and gelation, one important application of chemorheologicał models is the ability to design resin chemistry and processing conditions in order to obtain specific viscosity characteristics. As an example, too low resin viscosity during autoclave curing of carbon fiber/epoxy may cause starvation problems. On the other hand, if the resin viscosity is too high, consolidation problems may arise. Then, the cured composite may contain voids and other defects. Therefore, a processing window in terms of required viscosities must be defined in order to produce a composite with the required resin content.

HEAT TRANSFER

As discussed earlier, the cure of a thermoset is always associated with a significant development of heat. The temperature distribution inside the composite will depend on the competition between heat generation and heat diffusion through the thickness. While for thin laminates it can be assumed that the predominant heat diffusion will favor isothermal conditions, for thicker laminates the conditions will clearly be non-isothermal. Then, a heat transfer model for the prediction of the temperature distribution in the composite during curing must be provided. Material properties, boundary and initial conditions and the thermokinetic behavior are the input data of the heat transfer model. Indeed, kinetic and heat transfer model are coupled and a numerical solution method must be used.

Several authors have described how the temperature inside the composite can be calculated [2,5,6]. The law of conservation of energy is used together with the expression for the cure kinetics. Energy transfer by convection is neglected and so the energy equation becomes

   (7)

where Г and C are the density and the specific heat of the composite, z is the coordinate perpendicular to the plane of the composite and k is the thermal conductivity in the direction of z. The rate of heat generated, dH/dt can be obtained from the thermokinetic model:

   (8)

Hr is the total heat developed by the cure reaction.

The solution of the heat transfer equation can be obtained once the initial and boundary conditions have been determined. The composite properties Г, Cp and k can be calculated from appropriate models [2,5]. The variations of these parameters with temperature do not have a strong influence on the result of the calculations [5].

Kenny et al [5] used a slightly different approach introducing dimensionless numbers to facilitate the numerical solution and to generalize the model to other processing technologies. For this purpose, a characteristic time of the material behavior is defined: the isothermal gel time, tg, representing the time interval in which the material is changing from the liquid to the rubbery state. Taking the imposed processing temperature Te as the reference temperature, tg is obtained by integrating the thermokinetic equation. Then the dimensionless variables of the model are defined as:

   (9)

where To is the initial temperature and h is the thickness of the composite. The isothermal gel time tg can be viewed as a yardstick of the natural time scale for the crosslinking reaction. Then, the ratio of such a time scale to the characteristic time for heat diffusion is here defined as a Deborah number:

   (10)

De in this case represents the relationship between the heat transferred by conduction and the accumulated heat in the material. The Stefan number is:

   (11)

and may be considered as the relationship between the latent heat associated with the curing reaction and the accumulated heat in the material. The final energy balance becomes:

   (12)

When dimensioniess numbers are used, since first order parameters are involved in the differential equations, problems associated with convergence and stability become small. But perhaps the most attractive advantage is that by varying the dimensioniess numbers effects of several parameters can be studied. In addition, the same model can easily be extended to model other processing technologies like pultrusion, RTM and filament winding.

RESIN FLOW

Springer [12] presented a model for resin flow considering both horizontal and vertical flow. More recently Kardos and coworkers [4,13] presented a three dimensional consolidation and resin flow model which predicts: a) resin pressure and velocity profiles inside the composite as a function of position and time; b) the consolidation profile of the laminate as a function of position and time; and c) the resin content profile as a function of position and time.

When pressure is applied to a composite bleeder system in an autoclave, resin flows from the composite in the directions normal and parallel to the laminate surface. The resin flow rate depends on: a) the magnitude of the portion of the applied pressure which is transmitted to the resin, b) the gas pressure in the bag, c) the viscosity of the resin, d) the specific permeability of the fiber network in each direction, e) the porosity of the fiber network, f) the dimensions of the composite laminate.

In reference [4] a model for resin flow was presented, using analogies from the consolidation and flow theory used in soil mechanics. A porous bed of fibers was considered to behave like a spring. Earlier models nave assumed that the resin bears all the pressure. In reference [13] a viscoelastic resin flow model was presented which predicts the direction of resin flow during cure and the pressure gradients. Nonlinear pressure gradients exist in both vertical and horizontal directions and are the cause of the resin flow and consolidation processes. The governing equations of this model are:

   (13)

   (14)

   (15)

where e is the resin/fiber volume ratio (e = Vr/(1-Vr)), Vr is the volume fraction of resin, Kz and Kx are the permeabilities of fiber bed in the z and x directions, Pr is the resin pressure and Vx, Vz are the local average flow velocities in the x and z directions. In order to solve equations (13-15), Kz and Kx have to be expressed as a function of Vr using for example the Kozeny-Carman relation for permeability.

VOID MODEL

The void model should predict the void size distribution as a function of time, temperature and pressure during prepreg cure. For the construction of this submodel a description of the thermokinetics, the viscosity and flow behavior of the resin is needed as input data. Unfortunately, it is very difficult to obtain all the information needed to perform the predictions. But a simplified void model can at least help to limit the risk for voids in our choice of the optimum cure conditions in terms of time, temperature and