Nonlinear Tensile and Shear Behavior of Macro Fiber Composite

Actuators
R. Brett Williams1, Marc R. Schultz, Michael W. Hyer, Daniel J. Inman and W. Keats Wilkie
1
Corresponding Author
Graduate Research Assistant,
Department of Mechanical Engineering
Virginia Tech
310 Durham Hall MC 0261
Blacksburg, VA 24061 [email protected]

ABSTRACT

The Macro Fiber Composite actuator, developed at the NASA Langley Research Center, offers
much higher flexibility and induced strain levels than its monolithic piezoceramic predecessors. This
increased performance results from a laminated, piezoceramic fiber-reinforced construction and an
interdigitated electrode pattern. Since the Macro Fiber Composite is capable of such high actuation
strain levels, (~2000 µε, peak-to-peak), a significant amount of nonlinear mechanical behavior
occurs. As a result, the mechanical properties are not constant over the entire applicable strain
range of the device, and such changes should be accounted for in structural actuation models. The
current work presents the results of MFC tensile tests designed to measure, under short circuit
boundary conditions, both the four independent linear elastic engineering constants, valid at most
strain levels, as well as nonlinear stress-strain behavior. In addition to experimental results, various
plastic deformation models are developed to represent this constitutive behavior at higher strain
levels.

Keywords: piezoceramic fibers, elastic properties, material nonlinearities

INTRODUCTION

The Macro Fiber Composite (MFC) actuator is part of an emerging technology that strives to
improve the current state of the art for structural actuation, which currently relies heavily upon
monolithic piezoceramic materials. The MFC is a layered, planar actuation device that employs
rectangular cross-section, unidirectional piezoceramic fibers (PZT) embedded in a thermosetting
polymer matrix. This active, fiber-reinforced layer is then sandwiched between copper-clad Kapton
film layers that have an etched interdigitated electrode pattern. During manufacturing, these layers
are laid-up by hand and then cured in a vacuum hot-press. After the epoxy matrix that bonds the
package together is fully cured, a high DC voltage (~1500 Volts) is applied to the electrodes, thereby
poling the piezoceramic material in the plane of the actuator and establishing the poling direction
parallel to the PZT fibers. This in-plane poling and subsequent in-plane voltage actuation allows the
MFC to utilize the d33 piezoelectric effect, which is much stronger than the d31 effect used by
traditional PZT actuators with through-the-thickness poling [1]. A detailed description of the MFC
manufacturing process is given by Williams et al. [2] and, while the current work focuses on room-
temperature properties only, predicted elastic behavior over a wide range of temperatures is presented
in Williams, Inman and Wilkie [3].
Since the MFC is composed of both isotropic and orthotropic layers, it is most adequately
represented as a symmetric, hybrid, cross-ply laminate, or [IsoKapton /Isoacrylic/90°copper/ 0 PZT]S. Using
standard 1-2 notation for the principal material coordinate system [4], the overall mechanical
behavior of this orthotropic laminate is fully described by four independent stiffness quantities, E1,
E2, ν12 and G12, where ν12 is the major Poisson’s ratio. In a previous work, micromechanical models
were used to determine the elastic constants of each orthotropic layer, and then classical lamination
theory was used to determine the effective mechanical properties of the MFC under short-circuit
boundary conditions [2]. The present effort serves to verify these linear values experimentally, and
then to further characterize plastic deformation behavior.

EXPERIMENTAL PROCEDURE

In order to characterize their tensile and shear constitutive behavior and mechanical properties,
three sets of tensile test experiments were performed on MFC actuators. The first set was designed to
measure the stress-strain behavior in the PZT fiber direction, while the second set captured such
behavior transverse to these fibers. The third set measured the in-plane shear stress-shear strain
performance. This section describes the specific preparation of the various types of specimens, strain
gage instrumentation, and Instron testing machine set-up.

Specimen Preparation

The first set of tests loaded the standard MFC tensile test specimen, seen in Figure 1, in the
direction of the PZT fibers (1-direction). The tabs near the ends of the specimen were for loading by
way of pneumatic grips, and a fine mesh screen was inserted between the specimen and the grips to
prevent slipping during the test. The second set of tests loaded the MFC perpendicular to the PZT
fibers using 2-direction test specimens, which were cut as depicted in Figure 1 and fitted with
specially designed fiberglass tabs as seen in Figure 2a. This particular region was selected because it
contained only an active portion of the MFC (area containing copper electrodes), along with the
solder tabs and leads which preserved the electrical connectivity of the device. Clearly, the 2-
direction test specimen was much shorter than those used for the 1-direction tests, and having enough
area to fit into the grips while uniformly distributing the clamping pressure was an issue that was
resolved by bonding fiberglass tabs onto the specimen using strain gage epoxy. Fiberglass was
chosen as its stiffness closely resembled the stiffness of the MFC in the 2-direction, however, the
increased thickness of the specimen due to the tabs required the use of mechanical instead of
pneumatic grips. The slots seen in Figure 2a were cut in order to allow the MFC leads to escape from
the grips and be twisted together without damage so that the desired short-circuit properties were
measured. The third set of tests utilized 45° off-axis tensile test specimens, where θ is the PZT fiber
orientation angle seen in Figure 2b, to measure the in-plane shear stress-shear strain behavior of the
MFC. Again, the end tabs were used for loading into the grips; however, off-axis tensile specimens
exhibit shear-extension coupling. Since traditional mechanical or pneumatic grips prohibited such
shear deformation, special pivoting grips [5] that allowed the ends of the specimen to rotate freely
were used to promote a uniform state of stress across the specimen width.

Strain Gage Instrumentation and Instron Testing Machine Setup

For the 1 and 2-direction tests, strain gages capable of measuring both longitudinal and transverse
strains simultaneously, type CEA-13-250WQ-350 from Vishay Measurements Group, were aligned
with the 1-2 coordinate system seen in Figure 1, and then bonded to the center of the top and bottom
of each MFC tests specimen. For the off-axis tests, 0°-45°-90° strain gage rosettes, type CEA-13-
250UR-350 from Vishay Measurements Group, were aligned with the x-y coordinate system, and
bonded to the top and bottom of the specimens as shown in Figure 2b. All strain gages were bonded
using M-Bond AE-15 strain gage epoxy, which was cured at 150°F for two hours as suggested by the
manufacturer, and strain gage leads attached using solder. Each MFC specimen was then tensile
Figure 1 MFC Tensile Test Specimen With 1-2 Coordinate System.

tested using an Instron 4468 universal testing machine with a self-calibrating 1000 kN load cell and
the appropriate grips. All specimens were loaded at a crosshead speed of 0.2 mm/min until failure
occurred in the specimen or near the grips. During the tensile tests, a ten channel stand-alone strain
conditioner-amplifier and an analog to digital converter were connected to a PC, on which a Labview
Virtual Instrument was used to collect numerous sets of data simultaneously, namely load from the
testing machine and four (or six, for off-axis test) channels of voltage from the four (or six) strain
gages.

ELASTIC-PLASTIC DEFORMATION MODELS

Most engineering materials exhibit a simple, proportional relationship between stress and strain and
between transverse and longitudinal strain at low levels of stress and strain. The elastic moduli and
Poisson’s ratios are extracted from these linear relationships. However, at higher stress/strain levels,
this linear relationship is no longer valid, as plastic deformation begins to occur. Since the MFC is
capable of actuating at such high levels, it is important to characterize both its elastic and plastic
behavior. This section focuses on experimentally characterizing such behavior using various plastic
deformation models. Also, as part of this analysis, the experimental engineering properties are
compared to the values obtained using classical lamination theory.

(a) (b)

Figure 2 (a) 2-Direction Test Specimen, with Fiberglass Tabs and (b) Off-Axis Tensile Specimen with PZT
Fibers at θ = 45°.
Comments on Data Analysis

For an engineering analysis, stress and strain values must be computed from the acquired load and
voltage data. The actual strains are easily calculated from the recorded voltages, as the strain gage
amplifier is calibrated such that one output volt equals 1000 microstrain (µε). The strain values are
then averaged from the top and bottom gages to remove any bending strain effects. The stress
developed from the applied load is calculated by dividing the load data by the cross-sectional area of
the active region of the MFC. However, the chosen area represents the maximum value for the
specimen, as the copper electrodes are spaced out, rather than covering the entire active region. Thus,
the actual thickness can vary slightly, depending on the amount of epoxy that is expelled from
between the electrodes during the final MFC curing phase. Nevertheless, a photomicrograph of the
MFC cross-section indicates that using this maximum thickness is a reasonable approach [1]. Lastly,
the stress-strain curves are shifted horizontally so that the linear region extrapolates exactly through
the origin in accordance with ASTM Standard D3039/D3039M-00.

Tests in the 1 and 2-Directions

A quick examination of the stress-strain data from the 1 and 2-direction tests shows distinct elastic
and plastic strain regions. Behavior of this type is commonly represented using two plastic
deformation models, namely the elastic-linear hardening and Ramberg-Osgood relationships. Also,
the use of biaxial strain gages for these tests allow transverse strain data to be collected, from which,
Poisson’s ratios can be calculated. However, only the ν12 is considered independent and of particular
importance to the current characterization of the MFC.

Elastic-Linear Hardening Relationship

The elastic-linear hardening relationship describes the stress-strain behavior as linear at a slope of
Ei (i = 1, 2) up to the onset of plastic deformation, and then linear again at a slope of δiEi through the
plastic region, where δi is the slope reduction factor for the i-direction test [6]. For materials
displaying hardening type plastic deformation, i.e., the MFC when tested in the 1 and 2-directions, δi
has a value between zero and one. Mathematically, this type of constitutive behavior is given as

E ε (σ i ≤ σ 0i )
σi =  i i (1)
(1 − δ i )σ 0i + δ i Ei ε i (σ i ≥ σ 0i )

where σi is the normal stress, εi is the normal strain, and σ0i is the stress at which plastic deformation
begins. Application of this model to the data from the 1 and 2-direction tests requires a linear
regression of both the elastic and plastic regions, which provides values for Ei and δiEi for each
specimen, respectively. The onset of yielding is determined graphically as the point where the stress-
strain curve begins to exhibit plastic deformation, in contrast to an offset yield point typically used
with engineering metals. These characteristic values are averaged from the four specimens for each
of the two directions and appear in Table 1. Figure 3a shows a plot of Eq. 1 using these parameters,
along with the experimental data. Clearly, the models adequately represent the experimental data
from the 1 and 2-direction tests.

Ramberg-Osgood Relationship

The Ramberg-Osgood plastic deformation model represents the total strain as the sum of the elastic
and plastic strains components [6]. Again, the elastic strain component is proportional to stress,
however, the relationship between the plastic strain component and stress is exponential and given by
 (a) (b)

Figure 3 (a) Elastic, Linear-Hardening Model and Experimental Data and (b) Ramberg-Osgood Model and
Experimental Data: 1 and 2-Direction Stress-Strain Relationships.

σ i = H i ε Pi n
i
(2)

where Hi is a material constant, ni is a strain hardening exponent, and εPi is the plastic strain
component. Solving Eq. 2 for εPi and adding to the elastic strain gives the total strain as

1
σ  σ n
ε i = i +  i 
i
(3)
Ei  H i 

From the exponential nature of Eq. 2, a plot of stress versus plastic strain is a straight line on a log-log
plot. Values of Hi and ni for a particular tensile test are found either graphically from this plot (where
Hi is equal to σi at a εPi equal to one, and ni is the slope of the line if the logarithmic decades in the
two directions are of equal lengths) or using a linear regression analysis of logarithmic data. The
current effort uses the more precise latter approach to find Hi and ni for each of the experimental
stress-strain relationships. The averaged results for both the 1 and the 2-direction tests are presented
in Table 1. Figure 3b plots Eq. 3 with these values, one curve for each direction, along with the
corresponding experimental data. Again, the curves closely match the data.

Nonlinear Behavior of the Major Poisson’s Ratio, ν12

The experimental transverse strain data is used to determine Poisson’s ratio, ν12. As seen in Figure
4a, the slope of the linear fit, ν12, is constant in the linear elastic region; however, at the onset of

Table 1 Elastic-Linear Hardening and Ramberg-Osgood Parameters for 1 and 2-Direction Behavior

Ei, δiEi, σ oi, Hi, Plastic

Test Type δi ni ν 12 ν 21
GPa GPa GPa GPA ν 12
1-Direction
29.4 3.53 0.120 0.035 0.0450 0.0545 0.312 -0.506 N/A
Test (i = 1)
2-Direction
15.2 1.95 0.129 0.028 0.172 0.102 N/A N/A 0.161
Test (i = 2)
 (a) (b)

Figure 4 (a) Minus Transverse Versus Longitudinal Strain for 1-Direction Test and (b) Ramberg-Osgood
Model, Quadratic Least-Squares Fit and Experimental Data for Shear Stress-Shear Strain Relationships.

plastic behavior, it changes sign and magnitude. These changes are indicative of broken fibers
returning to a lower or stress-free state, at which time the transverse strain is recovered. Beyond this
strain level, ν12 is essentially zero. Poisson’s ratio values for the elastic and plastic regions are given
in Table 1. It should be noted that the tension tests from this research effort are monotonic, and fiber
breakage is a one-time event. Thus, Poisson’s ratio could change for subsequent tensile loading and
unloading scenarios, as would be the case for actuating the MFC under large harmonic voltage (above
~1250 volts peak to peak).

Comments on the Minor Poisson’s Ratio, ν21

Since the minor Poisson’s ratio, ν21, is not an independent material property, it is not essential to
verify its value experimentally. However, if one wishes to do so, it is determined in the same manner
as ν12, except that data from the 2-direction test is substituted for the 1-direction data. In the current
experiments, transverse strain data was collected for the 2-direction tests, however, since these test
specimens are so short, the fiberglass tabs severely restrict the transverse contraction, thus resulting in
experimental values of ν21 that are much too low. The results presented in Table 1 and 3 are from
calculated from the reciprocity relationship

E2
ν 21 = ν 12 (4)
E1

In Tables 1 and 3, the quantities on the right-hand side of Eq. (4) are from experimental and analytical
values, respectively.

Off-Axis Tests
While the 1 and 2-direction tests create a uniaxial state of stress by applying the load at 0° or 90° to
the PZT fibers, respectively, loading at any intermediate angle creates a biaxial stress-state consisting
of longitudinal, transverse and in-plane shear stresses on the off-axis plane [7]. Since this set of tests
is concerned with characterizing shear behavior, the load should be applied at an angle which
maximizes the shear strain in the material coordinate system. While previous work shows this
optimal angle is 32° for the MFC [7], special tensile-test coupons were manufactured with the PZT
fibers oriented at 45° because such components are readily available from NASA Langley. However,
this deviation in angle results in only a 5% reduction in shear strain [8] and simplifies the subsequent
data analysis.
In order to obtain a shear stress-shear strain curve, the collected stress-strain data from an off-axis
specimen must be manipulated in the following manner. First, stress transformations give the shear
stress in the material coordinates as

σ x sin 2θ
τ 12 = (5)
2

where σx is the applied unidirectional stress and θ is the PZT orientation angle defined in Figure 2b.
Next, the shear strain in the material coordinate system is

γ 12 = (ε y − ε x )sin 2θ + γ xy cos 2θ . (6)

Here, for 0°-45°-90° rosettes, εx and εy are equal to the strains from gages 1 and 3 (εg1 and εg3),
respectively, while the shear strain in the global geometric coordinates, γxy, is

γ xy = −ε g1 + 2ε g 2 − ε g 3 . (7)

However, since θ = 45° for these MFC coupons, Eq. 5 simplifies to

σx
τ 12 = (8)
2

and γxy is not required as the last term in Eq. 6 vanishes. Now, the desired relationship is found by
plotting the shear stress from Eq. 8 versus the shear strain calculated in Eq. 6 at each load level. The
in-plane shear modulus, G12, is the slope of the initial tangent to this curve.

Ramberg-Osgood Relationship

The Ramberg-Osgood model developed above easily adapts for use with nonlinear in-plane shear
deformation by substituting shear stress and strain parameters for their normal stress and strain
counterparts. The resulting expression for the intralaminar shear stress, τ12, is

τ 12 = H12γ P12 n 12
(9)

where H12 is a material constant, n12 is a strain hardening exponent, and γP12 is the plastic component
of shear strain. Solving Eq. 9 for γP12 and adding it to the elastic strain component gives the total
strain as

1
τ  τ n
γ 12
12
= 12 +  12  (10)
G12  H 12 

Following the regression analysis described above, average values for H12 and n12 are found from five
test specimens and the corresponding shear stress-shear strain relationship from Eq. 10 is plotted
along with the experimental data in Figure 4b.
Quadratic Least-Squares Fit

Clearly, the Ramberg-Osgood model accurately represents the shear stress-shear strain data.
However, it is preferable to have two expressions for all stress-strain curves, at least one of which can
be solved explicitly for stress. Examination of the data shows only a small amount of plastic strain,
so a power hardening type of relationship will not fit the data well. However, the shear stress-shear
strain data does appear to possess a quadratic relationship of the form

τ 12 = C1γ 12 + C 2γ 122 (11)

where the constants C1 and C2 are found using a quadratic regression procedure for each specimen.
The average values from five tests are presented in Table 2 and are used to plot Eq. 11 along with the
experimental data in Figure 4b.
Unlike the elastic-linear hardening and Ramberg-Osgood models, this quadratic regression is
purely mathematical and not based on any physical behavior. However, Eq. 11 provides a good fit to
the data and gives a simple shear stress prediction that can be readily applied for dynamic simulation
and control models.

Comparison with Theoretical Linear Values

In a previous work [2], classical lamination theory was used to predict effective laminate stiffness
properties of the MFC. These values are presented in Table 3, along with the percent error when
compared to the experimental values obtained from the current effort. Clearly, the traditional
mechanics of composite materials theories predict, with reasonable accuracy, the four independent
engineering properties in the linear elastic region. This experimental verification of these well-known
equations is significant because future researchers can employ such techniques for alternative MFC
lamination configurations, including various PZT fiber orientations and geometries.

CONCLUSIONS

This research effort experimentally determines the four independent orthotropic engineering
properties in the linear elastic region as well as characterizes the nonlinear constitutive behavior of
the MFC actuator under short-circuit conditions. The experimental values of E1, E2, ν12 and G12 agree
with published values [2] to within a few percent in most cases. Therefore, a researcher can use
traditional mechanics of laminated composites to develop new types of MFC actuators with various
piezoelectric materials and geometries. Also, the models in Figure 3 depict high fidelity quantitative
relationships for the nonlinear stress-strain behavior in the 1 and 2-directions, while the off-axis test
specimens provide accurate representation of the in-plane shear behavior of the MFC, as shown in
Figure 4b. However, these models all have both advantages and disadvantages. For the 1 and 2-
direction tests, one could argue that the Ramberg-Osgood is a better fit overall, particularly in the
elastic-plastic transition region, yet it involves increased complexity, the inability to explicitly solve
for stress, and no distinct yield point. While in general, either model is acceptable for design use, one
concerned with overall stress-strain behavior would likely prefer the simpler elastic-linear hardening
model, but if a failure and durability study of the MFC is performed, the Ramberg-Osgood model
would be preferable

Table 2 Ramberg-Osgood and Quadratic Regression Parameters for Shear Behavior

G 12 , GPa n 12 H 12 , GPa C 1 , GPa C 2 , GPa

6.06 0.289 0.138 6.20 -384

Table 3 Predicted Orthotropic Elastic Constants From Classical Lamination Theory

Ei, Ei % ν 12 % ν 21 % G 12 , G 12 %
Test Type ν 12 ν 21
GPa Error Error Error GPa Error
1-Direction
31.13 5.9% 0.254 -18.5% N/A N/A
Test (i = 1)
7.00 15.5%
2-Direction
18.55 22.3% N/A N/A 0.1514 -5.84%
Test (i = 2)

due to its fidelity in the critical yielding region. As for the shear stress-shear strain behavior, both the
Ramberg-Osgood model and the quadratic regression fit the data extremely well. While the former
approach has the aforementioned drawbacks, it is derived from observed mechanical behavior, while
the latter is simply a mathematical formula with no physical basis. Also, the fit lines in Figure 4a, the
slopes of which represent the major Poisson’s ratio, represent the data well over the entire applicable
strain range. Overall, this project gives researchers in the field of intelligent structures an in-depth
view of the short-circuit mechanical behavior of the MFC actuator, including both values for essential
elastic engineering constants as well as equations designed to handle even complex plastic
deformation.

ACKNOWLEDGEMENTS

The financial and technological support of the following organizations is gratefully acknowledged:
-Air Force Office of Scientific Research (AFOSR), grant number F49620-99-1-0231
-NASA Langley Research Center, grant number NASA LaRC 01-1103
-NASA Graduate Student Researchers Program (GSRP), grant number NGT-1-02014
-NASA Virginia Space Grant Consortium

REFERENCES
1. Wilkie, W. K., Bryant, G. R., High, J. W. et al., “Low-Cost Piezocomposite Actuator for Structural Control
Applications,” Proceedings, SPIE 7th Annual International Symposium on Smart Structures and Materials,
Newport Beach, CA, March 5-9, 2000.
2. Williams, R. B., Grimsley, B. W., Inman, D. J. and Wilkie, W. K., "Manufacturing and Mechanics-Based
Characterization of Macro Fiber Composite Actuators," in proceedings of 2002 ASME International
Adaptive Structures and Materials Systems Symposium, New Orleans, LA, November 17-22, 2002.
3. Williams, R.B., Inman, D.J. and Wilkie, W. K., “Temperature-Dependent Coefficients of Thermal
Expansion for Macro Fiber Composite Actuators,” Proceedings, 5th International Congress on Thermal
Stresses, Blacksburg, VA, June 8-11, 2003.
4. Hyer, M.W. Stress Analysis of Fiber-Reinforced Composite Materials, WCB/McGraw Hill, New York,
1998.
5. Pindera, M. J. and Herakovich, C. T.“Shear Characterization of Unidirectional Composites with the Off-
Axis Tension Test,” Experimental Mechanics, March 1986, pp. 103-112.
6. Dowling, N. E., Mechanical Behavior of Materials, Second Edition, Prentice Hall, New Jersey, 1999.
7. Chamis, C. C. and Sinclair, J. H., “Ten-deg Off-Axis Test for Shear Properties in Fiber Composites,”
Experimental Mechanics, pp 339-346, Sep. 1977.
8. Williams, R. B., “Nonlinear Mechanical and Actuation Characterization of Piezoceramic Fiber
Composites,” Virginia Tech Ph.D. Dissertation, September, 2003.