Processing and Performance of Composites with Micro and Nanoscale Reinforcements Richard J. Johnson Major Advisor: Associate Advisors:
Ranga Pitchumani Richard Parnas Kenneth Reifsnider Michael Renfro Jiong Tang
University of Connecticut Advanced Materials and Technologies Laboratory 191 Auditorium Road Storrs, CT 06269 www.engr.uconn.edu/amtl Dissertation Defense, Storrs CT, January 3, 2007
Outline The dissertation addresses three issues pertaining to the processing and properties of thermosettingmatrix composites with micro- and nano-scale reinforcements.
1. Active control of a liquid molding process 2. Characterization of chemorheology and cure kinetics of an epoxy resin system with carbon nanotube fillers 3. Carbon nanotube based damping of glass-epoxy composites
Process Description and Problem Motivation • Vacuum Assisted Resin Transfer Molding (VARTM)
• Preform permeation is a critical step – voids and dry spots = poor part quality • Flow control that eliminates the defects is essential for robust processing
Existing Control Strategies • Manipulate pressure or flow rate at inlet or vents to steer the flow in a desired pattern through the preform
D. Nielsen, R. Pitchumani (COMPOS PART A-APPL S: 2001) (COMPOS SCI TECHNOL: 2002) (POLYM COMPOSITE: 2002)
• Boundary control methods show reduced controllability as the flow moves farther from the controlled boundary
Proposed Active Control Strategy • Fundamentally, the flow through the fibrous preform is governed by Darcy’s Law for flow through permeable material Preform permeability tensor Volume averaged flow velocity
v P
Volume averaged pressure gradient Resin viscosity
• Local flow velocity can be changed by: – Option 1: Manipulating the local pressure gradient at the flow front by adjusting the inlet flow rate or pressure. However, the ability to target a pressure gradient change at a particular location on the flow front diminishes as the front moves farther away from the controlled boundary. – Option 2: Changing the viscosity of the flowing resin locally at locations of poor permeation in real time with the process. Altering the resin property in real time at targeted locations is referred to as active control.
• Proposed active control: Active localized induction heating – T → ; however, T → cure reaction rate → – Active control needs to balance these competing effects
Experimental Setup
Induction Heating Io L dl r H 4 0 r 3 emf 2f H z dA
Induction coil geometry
I2 I1
emf Electromotive force [V ] f Frequency[ Hz ] H Magneticint ensity[T ] A Area [cm 2 ] r Vector from gridi , j to dl [cm]
I3
I4
• Induction power calculation
o Tm Permeability of free space 4 A I Current[ A] dl Incramental coil length[cm] R Re sis tan ce []
emfi
– current conservation at the nodes of the susceptor mesh – summation of voltages in a loop = emf emfi I j R 2
Induced power
Ij R dA
Computational Process Modeling A 3-D nonisothermal model was developed for mold filling with reactive resin, accounting for the effects of induction heating •
Continuity and Momentum Eq. (Darcy’s law):
•
Energy Conservation: c
T T T T T T d k x k z a1 cu cv k y a2 q a1C A0 H r t x y x x y y z z dt
Cure Reaction:
•
Chemorheology:
•
Preform Compaction/Permeability: f P
E
d d d n u v K 0 e RT m 1 dt dx dy
•
•
Pf Pf 0 vf
Ae
E B RT
3 Cz 1 2
Coupled system of equations must be solved iteratively
Flow Enhancement with Induction Heating Validation 100
Illustrative Trends
Line Vent
–77 kPa 0V
80
% Fill
60 40 20 0 100
Experimental Simulated Line Vent Line Inlet
80
–77 kPa 140 V
% Fill
60 40 20
Experimental Simulated
0
Line Inlet
Results indicate that induction heating is capable of yielding sufficient flow enhancement without exceeding critical temperature limits. Operating windows serve as guidelines in implementation of active control.
Example Processing windows
Active Control of Unreactive Systems Goal: To build upon the flow enhancement results obtained from fixed coil positions to implement active control with appropriate coil motion. Coil motion and coil voltage control strategies are determined, focusing on unreactive systems first. Unreactive systems also approximate slow curing systems.
•
Controller Architecture: The controller’s objective is to (a) steer the flow in a uniform manner while (b) keeping the maximum temperature below a prescribed value (chosen to be 100º C) so as to prevent premature gelation, if a reactive resin system were to be used.
•
The coil voltage and coil location need to be estimated in real time toward these objectives. Numerical simulation studies on the active control were performed to determine the appropriate strategies for voltage and location estimation.
coil
i
ycoil(t+∆ti)
Lumped Capacitance Energy Model
y x
T(x,y,t*), Y(x,t-∆t ) i-1 xcoil(t), ycoil(t), Vcoil(t))
VARTM Setup
LabVIEW Y(x,t)
Vcoil(t+∆ti)
Coil Voltage Estimator
(xcoil(t+∆ti), ycoil(t+∆ti))
Induction Coil Vcoil(t+∆ti) On a x-y Stage x (t+∆t )
T(x,y,t*+∆ti-1)
•
Coil Location
Simulations of Active Control •
Simulations of the active control were conducted by replacing the actual process by the process simulation model. Several preform layups were studied to determine the coil y-motion strategy and the residence time y-motion Strategies
Preform Layups Studied
Line Vent
Ideal Uniform Flow Front
Flow Front
i
RMS
1
X
N 2 i i 1 N
Line Inlet
N
Figure 7, Johnson & Pitchumani
“Flow front following” scheme is a simpler strategy, provides better fill uniformity, and is chosen for implementation
Residence Time Determination •
Using the flow front following scheme, active control simulations were performed for a range of residence (delay) times: 0 – 15 s.
• • •
Optimum residence times range from 2-10 s. Minimum transient time for the induction coil used to go from 0–200 V is 7 s. A residence time of 10 s was chosen for the implementation
Active Controller Implementation Uncontrolled
Random Woven Mat Mat
Solid Insert
Controlled
With active control: fill time is greatly reduced, and fill uniformity is enhanced ending error is greatly reduced and there’s less wasted resin void entrapment is eliminated
Active Control of Reactive Systems Goal: Demonstrate the application of the active flow control on a reactive resin system Controller Objectives: (1) Minimize fill time (2) Improve flow front uniformity (3) Keep the degree of cure within a prescribed upper bound (x (t+∆t ), y (t+∆t ))
y
Real-time Threedimensional Thermochemical Model
x
t
VARTM Setup
LabVIEW
Y(x,t)
Simulation Time Real Time
i 1
i2
1sec .
ti 1
t1
t 2
ti 1
coil
i
Vcoil(t+∆ti) Coil Voltage Estimator
T*max(x,y,t+∆ti) Target Temperature Estimator t*fill(x,t)
((x,y,z,t*), Y(x,t-∆t ) i-1 T(x,y,z,t*), xcoil(t), ycoil(t), Vcoil(t))
i
(xcoil(t+∆ti), ycoil(t+∆ti))
Vcoil(t+∆ti) xcoil(t+∆ti) ycoil(t+∆ti)
(x,y,t*+∆ti-1)
Induction Coil On a x-y Stage
T(x,y,t*+∆ti-1)
coil
End of Fill Estimator
Coil Location
Simulations of Flow Control Goal: Investigate the performance of the controller and the effect of the constraint, L Preform Layups Studied Limiting Case
•
Relaxed cure constraint → Increased control action Critical cure constraint well satisfied.
Experimental Implementation Uncontrolled
Active Controlled
Active Control with 0/200V Setting
• Overall, localized heating based active control was demonstrated to enable reliable fabrication by eliminating fill related defects.
• The control scheme was also shown to be independent of preform geometry.
Outline Active control of a liquid molding process
Composites Science and Technology, 63, pp. 2201–2215, 2003 Composites A: Applied Sci. and Manufacturing, 37(10), pp. 1815–30, 2006 Composites Science and Technology, 67, 2007, In Press, DOI: 10.1016/j.compscitech.2006.04.012 Journal of Composite Materials, 2007 (In Review)
2. Characterization of chemorheology and cure kinetics of an epoxy resin system with carbon nanotube fillers 3. Carbon nanotube based damping of glass-epoxy composites
Introduction
Motivation: –
Carbon nanotubes (CNs) are being used in applications for improved properties: increased strength, damping, thermal and electrical conductivity; decreased thermal expansion – Processing of composites with carbon nanotubes requires knowledge of chemorheology and cure kinetics – Fundamental questions are whether the carbon nanotubes promote or hinder cure reaction – Despite its significance, a systematic quantitative study of the effects of carbon nanotubes and their parameters on the cure kinetics and chemorheology is not available in the literature Goals: – Characterize the effects of carbon nanotubes on the cure kinetics of EPON 815C epoxy catalyzed with EPICURE 3274 – Characterize the effects of carbon nanotubes on the chemorheology of carbon nanotube filled EPON/EPICURE resin system
Cure Kinetics Characterization Measure Resin / CN
Ultrasonic Processing
Degas
Mix/Crimp Sample
Measure Heat Flow
Differential Scanning Calorimetric (DSC) measurements were made to determine the thermal response of a resin specimen as compared with a standard when the two are heated uniformly. A comprehensive study enumerating 95 different experiments was conducted over a range of carbon nanotube morphology (MWCN, SWCN, and PSWCN), loading (0, 0.1, 0.3, 0.5 wt %), and aspect ratio
Weigh Sample
Thermograms of heat flow with time
Name
Diameter [nm]
Length [m]
Nominal Aspect Ratio
Aspect Ratio Range
30(1-5)
30 ± 15
1-5
100
22-333
15(1-5)
15 ± 5
1-5
200
50-500
30(5-20)
30 ± 15
5-20
416
111-1333
15(5-20)
15 ± 5
5-20
833
250-2000
Cure Kinetics Model For each sample, the measured thermograms under a nonisothermal (dynamic) DSC scan were used to determine the heat of the reaction For each sample, the measured thermograms under various isothermal temperatures (80–110 deg. C) were used to relate the cure reaction rate as a function of temperature and degree of cure, using an Arrhenius form: d K0 eE/ RT m 1 n dt
K
Incr. K with T
Heat flow–Time → Cure rate–Cure
Effect of Carbon Nanotubes on Cure Kinetics •
The heat of the reaction had no trend with CN morphology, loading or aspect ratio, and was taken to be average of all the measurements as 349.6 J/g
•
The reaction rate, K, exhibited no discernible trend with the morphology or the aspect ratio
•
The reaction rate, K, increases with carbon nanotube weight fraction, ω, and the increase was greater at the higher temperatures.
•
The effects of carbon nanotubes were incorporated in the cure kinetics model through a reduction in the activation energy with ω as d E 1 K0 e dt
2.5
RT m
1
n
K
K0 = 6.83x10-6 1/s; E = -64.67 KJ/mol; m =2.07 ; n =1.793
•
The decrease of E with ω is also consistent with the qualitative observation by Ton-That et al. 2004.
> 95% of data lie within 15%
Chemorheological Characterization Measure Resin / CN
Ultrasonic Processing
•Need measurements as a function of T , , and but can’t control cure.
Degas
Mix sample & place in viscometer
T , , and
•76 Experiments •Viscosity range changes with cure (time) •RPM (shear rate) must be dynamically selected to maintain the load range of the machine
Measure Viscosity
Chemorheological Modeling • •
Isothermal viscosity tests over range of machine: • • E B • RT n 1 General form: Ae
↑Temperature ↓Viscosity ↑Cure ↑Viscosity ↑Shear rate ↓Viscosity
•
Shear dependence determined at shear rate discontinuity.
•
Temperature and cure dependence found through successive least squared regression
–
Shear thinning
Effect of Carbon Nanotubes on Chemorheology
90% of data falls within 15% error band
Carbon nanotube dependence form modeled after the Mooney equation (Nielsen and Landel, 1994)
K E B 30.3log10 1 1 RT
A exp
A=6.16x10-11Pa·s; Eη=55.947KJ/mol; B=3.76; K η=170 MWCN, 148 SWCN
3wt.% 15nm diameter 1-5 um length MWCN
•
Results show a good fit for: – – –
Cure up to 0.4 Shear rate dependence Temperature dependence
Outline Active control of a liquid molding process
Composites Science and Technology, 63, pp. 2201–2215, 2003 Composites A: Applied Sci. and Manufacturing, 37(10), pp. 1815–30, 2006 Composites Science and Technology, 67, 2007, In Press, DOI: 10.1016/j.compscitech.2006.04.012 Journal of Composite Materials, In Review, 2007
Characterization of chemorheology and cure kinetics of an epoxy resin system with carbon nanotube fillers Rheologica Acta, In Review, 2006
3. Carbon nanotube based damping of glass-epoxy composites Composites Science and Technology, In Review, 2006
Introduction to CN-Composite Damping Motivation: Commonly viscoelastic coatings are used to increase structural damping – Drawback: Heavy, reduced performance at elevated temperature CN damping offers a light weight alternative that increases performance with temperature Goal: Quantify the effects of CN’s on damping in multiscale reinforced composites Experimental study: Study CN damping effects through variation of CN aspect ratio as well as morphology and weight percent in epoxy composite samples with and without fiberglass reinforcement.
Concept:
•
Weak interfacial bond between CN’s and the matrix material
•
Large surface area to mass ratio
Samples: • 3 MWCN aspect ratio samples • 1 SWCN sample • 1 and 2 wt. % Type
Diameter [nm]
Length [m]
Nominal Aspect Ratio
Surface Area: Weight Ratio [m2/g]
MWCN
30 ± 15
1-5
100
74
MWCN
30 ± 15
5-20
416
74
MWCN
15 ± 5
5-20
833
148
SWCN
~1.3
~10
7692
1709
Sample Preparation •
Rectangular composite plates with CN fabricated using compression molding from which test specimens were machined
Piston Heated and Cooled Platens
Fiberglass Reinforcement
Experiments Stiffness: ASTM 683 • • •
5 Experiments / dog bone 2 dog bones per sample type First 20 data points used to compute slope (Modulus)
Free vibration with initial tip deflection • • •
5 tests / sample type / tip displacement → 150 Experiments 10 sample types 3 tip displacements
•
Consistent release using a caliper clamped in a vice
INSTRON 5869
Data Reduction of Free Vibration Test Example Output
Rectified Curve
Convert: Voltage → Acceleration → Position Envelope Bt c 1 Model: X (t ) A e sin n t
eff
Bt
n t c
n
Effect of Variation of Glass Fiber Content •
Nonlinear strain dependence
•
↑ damping with addition of carbon nanotubes
•
↓ damping with addition of glass fiber reinforcement
•
Shows the combined effect of coulomb damping and viscous damping
•
CN’s should only effect coulomb damping so it is desired to separate the effects
4 in. beam length 15 mm initial tip deflection
Components of Effective Damping •
Decrement method suggested by B. F. Feeny (1996) Viscous damping: –
From the solution to the differential equation for a mass and spring with viscous and coulomb damping: X i 1 X i 1 e X i X i 2
1 2
Coulomb damping:
Period
X i e X i 1 (1) i 1 e 1 x k X i-1
X i X i 1 (1) i 1 2 x k
–
Evaluated at successive half periods of the fitted equation: X (t ) A e Bt c1
Rectified Position .
•
Xi
Tim e
Spacific damping capacity
dissipated kinetic energy cycle total kinetic energy cycle
Viscous and Coulomb Damping Effects Viscous damping •
No variation in viscous damping with carbon nanotube content
•
Increased coulomb damping with CN content and A/W ratio Coulomb damping A/W=74 m2/g
A/W=148 m2/g
67% Within experimental error
A/W=1709 m2/g
100%
Present Contributions / Future Work
Developed a novel active control strategy based on manipulating viscosity in real time, and demonstrated its effectiveness in enabling defect-free mold filling in VARTM for a variety of preform geometries.
For the first time developed chemorheological and cure kinetics models that quantify the effect of CN fillers.
Implementation of the active control methodology on complex threedimensional geometries could be explored in a future work.
Future study could focus more closely on the effect of aspect ratio on the rheology, including the effects of ultrasonic processing conditions on CN breakage (and in turn, aspect ratio alteration).
Measured for the first time CN induced coulomb damping in fiber reinforced composite samples Modeling to determine a true physical reason for CN damping and its reduction in the presence of micro scale fibers. Further work can be done exploring the frequency dependence of CN based damping using forced vibration or uniaxial tension testing.
Thank You Major Advisor Ranga Pitchumani Associate Advisors Jiong Tang Kenneth Reifsnider Michael Renfro Richard Parnas Funding Acknowledgments Owens Corning National Science Foundation (CTS-9912093, CTS-0522933) Air Force Office of Scientific Research (F-49620-01-1-0521)
Questions
Cure-Based Heating Control •
Control Principle: – –
•
Dictate a maximum cure at the end of fill, L Determine coil voltages that achieve the cure constraint at the end of fill
Step 1: Fill Time Estimation – Using Darcy’s law: v P –
Extrapolate the current flow front position and velocity to estimate the fill time.
–
Estimation performed in 1D along a line to get a fill time distribution across the mold.
L2 y o ( x, t ) 2 t fill ( x, t ) t t 2 y ( x , t ) ( y ( x , t ) y ( x , t t )) o o o
Indicates current measured value
L y
Cure-Based Heating Control • Step 2: Determine Limiting Temperature – Use the resin cure relationship:
d K 0 e E / RT m 1 n dt
– Determine the temperature that will lead to L at the end of fill – Assume • a second order reaction • Isothermal temperature
m n 2 (Commonly observed) T Tmax f (t )
Degree of Cure
Tmax ( x, y, t i ) L
o 0
0
t
t fill
Time
E R ln K 0 (t *fill ( x, t ) t )
L
o
m 1 2m d
Cure-Based Heating Control • Step 3: Determine Limiting Coil Voltage c
T ( x, y, t ) c f h T ( x, y, t ) T q( x, y, t ) t d c f 0.45
•
Cf is a factor that compensates for the assumptions of: – –
Negligible advection Uniform through thickness temperatures
c hti 1 T ( x, y, t ) T c f h Tmax ( x, y, t ) T exp f dc q ' ' ' ( x, y , t ) c hti 1 d 1 exp f d c
Coil Voltage Estimation • Determine Coil Voltage that does not exceed Tmax=100ºC c
T ( x, y, t ) c f h T ( x, y, t ) T q( x, y, t ) t d c f 0.45
•
Cf is a factor that compensates for the assumptions of: – –
Negligible advection Uniform through thickness temperatures
c hti 1 T ( x, y, t ) T c f h Tmax ( x, y, t ) T exp f dc q ' ' ' ( x, y , t ) c hti 1 d 1 exp f d c
Numerical Modeling - Flow • Flow governed by Darcy’s law: v • Pressure distribution: P 0
f
Pf
f
– five-point Laplacian scheme – Darcy’s law used to find velocities – volume tracking method used to find the flow front locations
• BC’s – Walls impenetrable with no slip – vacuum line defined with negative pressure – inlet defined by atmospheric pressure at the surface of the source container
• Permeability – Carman-Kozeny relationship: 3 Cz 1 2
Numerical Modeling - Heat Transfer • Energy equation: c
T T T T T T kz cu cv kx ky q t x y x x y y z z
• 3-D control volume analysis and ADI method (Douglas and Gunn: 1964)
• BC’s – mold sides considered adiabatic – top surface of the vacuum bag and bottom surface of the mold considered convective – inlet and outlet at ambient temperature
Numerical Modeling • Coupled by viscosity
6291.2 T
4.67 1010 exp
– Arrhenius equation: – flow is dependent on temperature through viscosity – temperature is dependent on the flow
• Iterative solution – convergence based on temperature:
• Time step varied – mesh Fourier number – mesh Courant numbers
t
T 105 Kelvin
.333 z 2 ut vt .333 .333 x y
Motion Control (Coil Location Estimation) • • •
Motion control involves positioning the coil appropriately at each time during the filling process so as to ensure overall fill uniformity. Three parameters need to be determined: x, y locations of the coil, and the residence time of the coil at each location. x-location: – Keep the coil behind the location of maximum lag. – To avoid undesired heating of other areas, the induction coil is turned off during its transit in the x-direction to the target location, and is supplied voltage only while above the desired location.
•
y-location: – Several strategies for coil location and motion in the y-direction are systematically investigated to determine the best strategy for implementation.
•
Residence (delay) Time: – Since the flow-front lag location changes dynamically during the process, a minimum residence (delay) time at each target location is needed to prevent the coil from leaving the target spot prematurely in chasing the maximum flow lag incessantly while never stopping to heat. – A range of residence time is systematically studied to determine the best values for implementation.