VIRTUAL ENGINEERING CENTRE

Prepared For: Julian Rishton AIRCELLE Client Reference: VEC Reference: Date: VEC.WP2.004.03 19 th January 2011 Issue 1

Prepared By: Approved By:

Dr G Abdelal Georgiou

&

Dr

Dr AJ Robotham Executive Director

Page 1 of 12

1.0 INTRODUCTION The Virtual Engineering Centre (VEC) is a University of Liverpool initiative in partnership with the Northwest Aerospace Alliance, the Science and Technology Facilities Council (Daresbury Laboratory), BAE Systems, Morson Projects and Airbus. The VEC provides a focal point for world class virtual engineering business, research, education and skills development, best practice demonstration and knowledge transfer for the aerospace sector. The VEC aims to fill a significant capability gap, particularly among small to medium sized enterprises in the aerospace sector, in the use of digital Virtual Prototypes to enhance business performance. An important role of the VEC is demonstrating the values of using virtual prototypes to companies so that they will effectively adopt them in the future, thus improving business performance. The VEC can offer a number of innovation support products and services, all based around the intelligent use of virtual prototypes to aid decision making, including; research and technology development, modelling and simulation services, training and continued professional development use of state of the art facilities. Funding support is available for eligible small to medium sized enterprises within the North West Region. Aircelle is.

2.0 BACKGROUND Autoclave processing is one of the oldest composites processing technologies. In this process, plies or prepregs or tapes (fibers that are preimpregnated with the resin) are stacked on the surface of the tool. They are then subjected to high pressure and temperature to allow the stacks to become a single coherent structure by forcing out air pockets and excess resin. The process is carried out in an autoclave, which is a large pressure vessel with an integral heating element. One can think of an autoclave as an oven that can be pressurized. Not unlike a pressure vessel, it is usually constructed as a cylindrical tube with a door at one end. Its main function is to provide the heat and pressure necessary to consolidate and cure composite structures.

Fig.1. Autoclave at Toyota Headquarter. The autoclave has been successfully used for many years to make composite parts which have no or negligible geometric complexity. Hence, the tool is usually a flat plate with very few curvatures and inserts. The tool material may be aluminium, steel, nickel, invar or even a composite depending on the application and the cost. A release film or a coating of release gel is usually applied on the tool surface for easy removal. The composite layup, consisting of prepregs or tapes, is placed on the tool surface sandwiched between release plies or fabrics. Cork or solid spacer plates are placed around the perimeter of the stacked plies, and they act as dams and limit resin flow in the lateral direction. The release film and release fabric are necessary for easy separation of the composite part from the tool surface and the vacuum bag. Bleeders are used to absorb excess resin in the thickness direction. 3.0 PROGRAMME OF WORK /ACTION PLAN The main objective of this project is the early prediction of composite part deformations during autoclave manufacturing. These deformation will help us to modify autoclave-tool design to minimize it within specified tolerances (customer requirements). Numerical autoclave simulation will reduce the cost of the iterative manufacturing process to find the optimized autoclave-tool design. Three software packages will be used during this project; CATIA, ABAQUS, and Fortran 11. CATIA is used to model the part and tool designs, ABAQUS is used to perform the Autoclave simulation applying finite element technique, and Fortran 11 is used for programming user-subroutines that model special material properties. Autoclave simulation consists from three components. The first component is the Thermo-Chemical model to simulate the curing cycle and determine the temperature and degree of cure distributions as a function of the cure cycle time. Second component is a Resin Flow Analysis to determine the final fiber volume fraction after removal of excess resin under autoclave pressure. Third component is the stress-thermal model to calculate residual stresses (that can affect part fatigue life) and deformation (that can affect assembly process). 3.1 THERMO-CHEMICAL MODEL

The following mathematical equation is modelled applying finite element technique (ABAQUS). A transient thermal analysis is performed that include element conductivity,

internal heat generated by epoxy during cure, and convection to the air in the Autoclave (convection term is not shown in the equation).

where denotes the composite density, cp the specific heat, T the temperature, t the time, xi the coordinates, kij the components of the thermal conductivity tensor, and c the degree of cure, which is defined as the ratio of the heat released by the reaction to the ultimate heat of reaction HR. Input parameters that are required for this model: 1. Fabric thermal conductivity tensor. 2. Cure kinetic model.

3.1.1 FABRIC THERMAL CONDUCTIVITY TENSOR

Fig. 2. Five-harness satin weave. (a) Unit cell; (b) fiber undulation in fill and warp direction. Based on Literature review: for in-plane thermal conductivity, closed analytical models are available but only for plain-weave, which is based on A micromechanics model that is based on the thermal-electrical analogy and it will take some time to extend it to 5harness satin weave. There is an analytical model for the satin weave to model the transverse thermal conductivity. The other method applied in this area, is a finite element model of the unit-cell of the 5-harness satin wave and run steady-state heat transfer analysis to predict the thermal conductivity tensor. We have two options to predict the thermal conductivity tensor; first option is the experimental data sheet from the fabric supplier. Second option is to predict the thermal conductivities of 5-harness satin weave composite, by modeling a unit cell using 3dimensional finite elements, considering the interlaced fiber tow architecture microscopically. At the unit cell boundary, the corresponding periodic boundary

conditions are applied. The model can be verified by the use the available analytical model prediction of the transverse thermal conductivity. If the finite element approach is used then the following parameters are required from Aircell: 1. 2. 3. 4. 5. 6. 7. Thermal conductivity of the resin and fibers. Width of the fill and warp strands (mm). Maximum thickness of the fill and warp strands (mm). Thickness of the unit cell (mm), which equals the lamina thickness. Strand volume fraction in the composite. Fiber volume fraction in the strand. Fiber volume fraction in the composite.

These parameters are required anyway, even if the thermal conductivity tensor is provided, as these parameters are needed for the prediction of the elasticity tensor, as will be shown later. 3.1.2 CURE KINETIC MODEL The last term in transient-thermal Equation corresponds to the internal heat generation associated with the exothermic cure reaction, and is described by a cure kinetics model

where g(T,c) represents some function of T and c, and takes on specific forms depending on the material system under consideration. This form require the definitions of 6constants, which can be available in literature or provided by the fabric supplier. These constants are most often determined using isothermal and/or dynamic DSC measurements and various curve-fitting techniques. The required experimental work, a total of eight additional DSC (differential scanning calorimetry) scans were performed. Of these, 6 were isothermal tests, two at each of 130 C, 150 C, and 170 C. In these tests, the specimens were heated rapidly to the desired temperature where they were maintained for a total of 2 hours, then rapidly cooled. Two dynamic DSC scans were also performed in which specimens were heated at rate of 10 C/minute from room temperature to about 290 C at which point cure was essentially complete. Many experimental data are already available in literature which can be used to estimate the values of these constants, and there is possibility that the composite centre at University of Liverpool will play a role regarding experimental work or data that are needed. 3.1.3 CURRENT PROGRESS We are able to predict the temperature and degree of cure distributions in a composite part, which is designed from Unidirectional Laminates. The model includes conductivity

of the part with autoclave-tool, and convection boundary condition to the surrounding air (boundary-elements are allowed to produce heat).

Autoclave-Tool

Composite Part

Temperature Distribution 3.2 RESIN FLOW ANALYSIS

Degree of Cure Distribution

As discussed in Hubert (1996), in the autoclave process flow of the matrix resin is induced to remove excess resin from the laminate, promote bonding between plies, and collapse any voids within the laminate. This flow process may also play an important role in the development of residual stress and deformation. Even in flat laminates, nonuniform resin distribution through the part thickness may develop due to an imperfect compaction process. V.A.F. Costa developed a three-dimensional numerical model developed to simulate and to analyze the mechanisms dealing with resin flow, heat transfer and the cure of thick composite laminates during autoclave processing. The model, which incorporates some of the best features of models already reported in the literature, is based on the Darcy law, the convection-diffusion heat equation, and appropriate constitutive relations. The same approach will be applied in our simulation, but instead of simulating the equations using finite-volume approach, the flow module in ABAQUS will be used. Input parameters required: 1. Fiber volume fraction in composite. 2. Resin viscosity model as a function of temperature and degree of cure. 3. Resin degree of cure at gelation.

3.2.1 CURRENT PROGRESS The process is not applied yet. This component will be added at a later step, after the third module is finished. 3.3 STRESS THERMAL MODEL During autoclave processing, a number of different processes lead ultimately to the development of residual stress and deformation. The mechanism for stress development in this case is the combination of these thermal strains with changing resin modulus. Unless both are present, no residual stress and deformation will be generated. From the mechanisms for the development of process-induced stress and deformation described in the literature, five main sources have been identified: Thermal strains (major). Resin cure shrinkage strains (minor). Gradients in component temperature and resin degree of cure (major). Resin pressure gradients (resulting in resin flow) (minor). Tooling mechanical constraints(major). Residual stress and deformations calculations applying sequential transient stressthermal analysis in ABAQUS, involves two main issues; Viscoelastic material model to simulate the resin behavior with temperature and degree of cure. Micromechanical model to predict the Elasticity tensor of the composite part with time as a function of the fiber and resin material properties.

3.3.1 VISCOELASTIC MATERIAL MODEL Characterization of the Viscoelastic response of a composite material is a complex problem, and is doubly difficult for a curing composite. The most serious difficulties arise from the need to measure material behavior over a wide range of temperatures and degrees of cure and the fact that curing of the resin changes its material response even as it is being tested. The amount and type of data collected, and the characterization method used depends on the Viscoelastic model chosen. The isotropic matrix resin in composite materials is modeled as a so-called curehardening/instantaneously linear elastic (CHILE) material. This designation indicates that the modulus of the instantaneously linear elastic resin increases monotonically with the progression of cure. Two models for prediction of resin modulus development are employed here, one from Bogetti and Gillespie (1992) and another developed by Johnston (1997). The estimation of the CHILE model parameters, requires searching literature for stressrelaxation test data of the specified epoxy (Hexcel material 914), or performing a set of

tests defined by Johnston (1997). The test in this case was performed using a Rheometrics RMS-800 rheometer on an 8-ply unidirectional specimen (fibers aligned with the twist axis) with approximate dimensions of 45 mm x 12.5 mm. The applied strain was varied from 1% to 10% with a frequency of 10 rad/s. The behavior of the resin after resin gelation and before complete cure, will be highly Viscoelastic. Any stresses generated during this stage will decay to some degree, although they may still be important to the final stress state. Based on this assumption and to not to increase the initial cost of the project, an approximate date from the literature will be used to the Hexcel-914 epoxy. Input parameters from Aircelle are: Fibers mechanical properties (Transverse Isotropic) [E11 - E22 - G13 - G23 - ]. Resin mechanical properties (isotropic) [E - ], for fully cured and at room temperature. Any material data at higher temperature or different degree of cure is welcome.

3.3.2 MICROMECHANICAL MODEL A micromechanical model called MESOTEX, which was developed by Universite de Technologie de Compiegne and Atelier Industriel de l'Aeronautique de Cuers Pierrefeu in FRANCE, will be used for prediction of the instantaneous elastic behavior of composites rein-forced satin weave fabrics. By using the classical thin laminate theory applied to each woven structure, this analytical model takes into account the strand undulations in the two directions and also integrates the geometrical and mechanical parameters of each constituent (resin, fill and warp strands). A representative volume is chosen for the woven composite and the fiber architecture is described by several functions. Input parameters required from Aircelle are: Fibers mechanical properties (Transverse Isotropic) [E11 - E22 - G12 - G23 - ]. Lamina parameters that already defined for the Thermal-Chemical model (section 3.1.1).

3.3.3 CURRENT PROGRESS A sequential stress-thermal analysis is performed based on the temperature and degree of cure predictions done on the Thermal-chemical analysis. Two approaches are applied to predict residual stresses and deformations: Explicit Finite Element Solver. Implicit Finite Element Solver.

A geometric with material nonlinear analysis requires incremental load steps. At the end of each increment the structure geometry changes and possibly the material is nonlinear or the material has yielded. Each of these things, geometry change or material change, may then need to be considered as you update your stiffness matrix for the next increment in the analysis.

An Explicit FEM analysis does the incremental procedure and at the end of each increment updates the stiffness matrix based on geometry changes and material changes. Then a new stiffness matrix is constructed and the next increment of load is applied to the system. In this type of analysis the hope is that if the increments are small enough the results will be accurate. One problem with this method is that you do need many small increments for good accuracy and it is time consuming. If the number of increments are not sufficient the solution tends to drift from the correct solution. An Implicit FEM analysis is the same as Explicit with the addition that after each increment the analysis does Newton-Raphson iterations to enforce equilibrium of the internal structure forces with the externally applied loads. The equilibrium is usually enforced to some user specified tolerance. So this is the primary difference between the two types of analysis, Implicit uses Newton-Raphson iterations to enforce equilibrium. This type of analysis tends to be more accurate and can take somewhat bigger increment steps, but Explicit FEM handle contact interaction better. 3.3.3.1 EXPLICIT FINITE ELEMENT SOLVER One of the major factors that affect the composite part deformations, is the interaction between part and autoclave-tool. One of the disadvantages that the Implicit solver suffers from is the contact iterations and specially if material nonlinearity is included in the analysis. That was the main reason behind the application of the Explicit FEM solver to predict the residual stresses and deformations. In Explicit FEM 3D solid elements can be used in the analysis but the composite solid section cannot be defined (software limitations). To resolve this issue, an averaging of the laminate elasticity tensor is performed in a MathCad sheet and then the calculated averaged elasticity tensor is used in ABAQUS to define orthotropic material behavior. Due to a BUG in the software that limits it from running Viscoelastic material model with User-Defined Shift function (this issue will be resolved next March), I run my sequential transient stress-thermal analysis starting from the point of fully cured up to the cooling the part down and tool removal step. Next, experimental results will be compared to the

numerical ones.

Fig. 3. Deformed-Unreformed Composite part after Tool Removal. 3.3.3.2 IMPLICIT FEM SOLVER Analysis results from an Implicit FEM solver are more accurate than an Explicit solver ones (if bigger time-step is used) due to the Newton-Raphson equilibrium iterations used. But due to the fact that Viscoelastic material model used in the Implicit FEM solver is isotropic one, which is not sufficient to model composite material. A HyperelasticOrthotropic material model is added to the Viscoelastic one to model the composite material more accurately. Adding contact interaction to the simulation causing it to be only stable at time-step increment of 10-7 (very slow). Currently, I am trying to model the CHILE model using the User-material subroutine in Abaqus (UMAT), which hopefully leads to faster running time. 3.4 DESIGN OPTIMIZATION AND SENSITIVEY ANLAYSIS Running iteratively the autoclave simulations to reach the optimized autoclave-tool design, is very costly (time wise). This long running time assuming that the part dimensions and material properties are fixed, and the only adjusted parameters are the tool dimensions. It will be even more costly, if optimization of the part material properties or laminate lay-up is required. To overcome this issue, a surrogate or neural-network approach (to be decided later) will be trained using the FEM results as an output, and part plus tool configurations as input. This scheme will make the design optimization more efficient and once an optimized solution is reached, it will be tested using the autoclave FEM simulation.

The autoclave numerical simulation is deterministic analysis, assuming all geometrical and material parameters are fixed at their mean value. But the manufacturing process is stochastic in nature, which means that the stresses and deformations can have normal or beta distributions, depending on the stochastic type of the input. Input parameters that have more influence on the final residual stresses and deformations are also required. So, running sensitivity and stochastic analysis will help us to find not only the mean value of the deformation or stresses but also how to narrow down their deviations (less standard deviation), using the more influential parameters. 3.2 VERIFICATION AND VALIDATION There are many available published data that can be used for verification of the process, but for unidirectional laminates. To be able to verify the simulation process of woven fabric we will need Aircelle will provide us with manufactured test part that can be used to verify the simulation process. There is no published data regarding the autoclave simulation of fabric laminates. 3.3 DELIVERABLES Include deliverables such as: Short summary report on modelling and simulation described above. Digital case study representation of results. 3.4 DECISION POINTS The initial programme of work involves 3 steps. Following Step 1, there is an opportunity to validate the simulation model against inspection data. If this comparison yields good agreement between simulated and measured results, then the programme of work should progress to Steps 2 and 3. If not, the programme of work will need to be re-evaluated. 4.0 SCOPE OF SUPPLY 4.1 SCOPE The scope of the work includes: 4.2 PRICE/STATE AID SUPPORT SUMMARY The above programme of work will be used to generate a case study that demonstrates the value of virtual engineering to a manufacturing process control problem. The case study will contribute to the deliverables of the publically funded VEC project with real life data being supplied by Airbus as a partner of the project. The VEC will have the authority to publish the generic findings of the programme of work within the limits of confidentiality agreed in 4.4.

4.3 EXCLUSIONS The programme of work is limited in the first instance to the evaluation of an idealised cell and idealised panel with specific material properties chosen from the 2xxx series aluminium alloys. The geometry of the idealised cell and panel excludes any specific reference to a production wing panel.

4.4 TERMS AND CONDITIONS This offer is subject to the standard terms and conditions of the partnership agreement between the University of Liverpool and Airbus for the VEC project. 4.5 TIMESCALE It is anticipated that the work outlined above will require 10 staff days of effort over 10 12 weeks and will be completed by 1st-April 2010.

5.0 DECLARATION I confirm that the above mentioned information is deemed correct and outlines the agreement made between the University of Liverpool and Airbus. For Aircelle Signed ________________________________ __________ Name ____________________________ Date

For the University of Liverpool Signed _____________________________ __________ __ Name ____________________________ Date