- Email: [email protected]
Design, manufacture and evaluation of bending behaviour of composite beams embedded with SMA wires
Composites Science and Technology 69 (2009) 2034–2041
Contents lists available at ScienceDirect
Composites Science and Technology journal homepage: www.elsevier.com/locate/compscitech
Design, manufacture and evaluation of bending behaviour of composite beams embedded with SMA wires Gang Zhou a,*, Peter Lloyd b a b
Department of Aeronautical and Automotive Engineering, Loughborough University, Loughborough, Leicestershire LE11 3TU, UK DSTL, The Ministry of Defence, Farnborough, Hants GU14 0LX, UK
a r t i c l e
i n f o
Article history: Received 15 July 2008 Received in revised form 15 January 2009 Accepted 18 January 2009 Available online 25 January 2009 Keywords: A. Smart materials B. Stress/strain B. Thermomechanical properties C. Deformation Shape memory effect
a b s t r a c t An experimental study has been conducted to design and fabricate smart composite beams embedded with prestrained nitinol wire actuators. The developed fabrication process allowed both quasi-isotropic E-glass/epoxy and carbon/epoxy hosts to be eccentrically embedded with 10 parallel prestrained wires with a purpose-made alignment device and cured successfully in an autoclave. Smart composite beams of three different lengths were made for each type of host. Both single-cycle and multi-cycle thermomechanical bending actuations of these beams in the cantilever set-up were characterised experimentally by applying various levels of electric current to the nitinol wires. The performance characteristics showed that the present fabrication process was repeatable and reliable. While the end deflections of up to 41 mm were easily achieved from smart E-glass/epoxy beams, the limited end deflections were observed from the smart carbon/epoxy beams due primarily to our inability to insulate the nitinol wires. Moreover, it seemed necessary to overheat the prestrained wires to much higher temperatures beyond the complete reverse transformation in order to generate recovery stress. The longer beams showed greater actuation rates and took less time to reach the same level of deflection. It was found that the actuation capability derived from single-cycle actuation exercises was not suited to multi-cycling actuations and could result in premature failure of multi-cycled smart beams. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction Shape memory alloys (SMAs) such as nitinol have the ability to exhibit diffusionless solid-state phase transformations, which are induced either by mechanical stress or by temperature. Commercially, they are available in wire or ribbon form. The two solid-state crystalline phases involved are room temperature martensite and high temperature parent phase austenite. Their unique transformations from austenite to martensite are characterised by four critical temperatures, i.e. martensitic start temperature Ms, martensitic finish temperature Mf, austenitic start temperature As and austenitic finish temperature Af. One type of transformation, called pseudoelasticity, superelasticity, or stress-induced martensite (SIM), occurs when SMAs in the isothermal austenitic condition are mechanically stressed into the plastic region so that their plastic strain is completely recovered on the removal of the stress. The other type of transformation, called the shape memory effect (SME), occurs when plastically deformed SMAs in the martensite state recovers their original shape (i.e. plastic strain), if heated electrically above their Af. If the recovery of the plastic strain is constrained, a large recovery stress will be generated. Since their * Corresponding author. Tel.: +44 01509 227 210; fax: +44 01509 227 275. E-mail address: [email protected] (G. Zhou). 0266-3538/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.compscitech.2009.01.017
discovery in the early 60s, a great deal of effort has been devoted to characterising SMAs, exploiting these two thermomechanical phenomena and developing new applications in aerospace, civil, mechanical and medical industries as discussed in [1–9]. In particular, as those transformation temperatures for any given SMAs can be tuned to suit a specific application via a change of the material composition and/or heat treatment, they can be developed into smart actuators through extensive exploitations of the recovery stress for smart adaptive structure technology. In addition to the aforementioned one-way SME, of current interest, SMAs could be trained to exhibit the two-way SME as reported in [10,11], but with a much limited recovery strain. To develop smart adaptive structural components by controlling their shapes in bending, SMA actuators in general need to be integrated with the host components such as composite structures, though some non-embedding approaches have also been investigated [1–4,9]. Important performance parameters for such smart components include those associated with SMA actuators such as maximum strain, prestrain, wire diameter, energy density ratio, stiffness, bandwidth (including cooling rate) and actuation capability; and those associated with the host such as actuator volume fraction, actuator through-the-thickness location, host flexural rigidity, actuator–host interfacial strength and actuation cycle. For the smart components to be adaptive, the respective contribu-
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
tions of the aforementioned parameters to their functional performance must be synergistic, which can be a significant challenge. This is because the functional attributes of their performance contributions sometimes lead to significant conflicting consequences, which compromise the performance of the components, as well as undesirable ‘side effects’. For example, an adaptive beam embedded with prestrained one-way SMA wires was expected nominally to show substantial deflection with a reasonable actuation rate, once energised through electric heating. Intuitively, it seemed desirable for the host composite beam to (a) have a large amount of highly prestrained wires embedded with a sufficiently large eccentricity to deliver sufficient bending moment and then substantial deflection; (b) have a relatively moderate flexural rigidity against which the beam bending induced by the not-so-high level of bending moment would have to overcome; and (c) be actuated by a high level of electric current to achieve desired actuation responsive rate. However, these attributes led immediately to several difficulties as they were conflicting requirements. Although nitinol wires had a recovery strain of up to 8%, the high level of prestraining could lead to their rapid ageing during multi-cycle actuations [12]. The thickness of the composite host remained the most difficult parameter to reconcile. One the one hand, a relatively large thickness offered a greater eccentricity and a relatively large flexural rigidity. The latter was required not only to return the actuated beam to its original position but also to prestrain the embedded wire actuators again for the next cycle. On the other hand, the greater flexural rigidity of the beam could become a greater obstacle for the host to overcome to achieve the desired bending. Moreover, as phase transformations were hysteretic and could thus release latent heat, a high level of applied current, desirable and necessary for a high actuation response rate, could result in heat damage in the composite host [12,13], which could in turn affect the wire actuation capability for subsequent cycles. In addition, embedding too many SMA wires in the host could also adversely affect the long-term through-thethickness mechanical properties of the host structures [14]. Although the literature abounds with individual endeavours of using SMAs for the shape control of smart adaptive structures as discussed in [1–9], a synergistic set of the performance attributes from the aforementioned parameters seemed clearly application specific involving inevitable compromises. In particular, the experimental thermomechanical behaviour of embedded SMA actuators, underpinning and justifying those compromises, has not been understood. To this end, it is of paramount importance to develop a good understanding of actuation characteristics of SMA-based adaptive composite beams in terms of performance bounds of the aforementioned important parameters. This experimental study aims on the examination of actuation characteristics of Eglass/epoxy and carbon/epoxy laminate beams embedded with nitinol wire actuators. Its focus will be on the manufacturability and actuation repeatability of adaptive composite beams and the effects of beam length and applied current level on their actuation characteristics. Information generated and experience gained will be channelled into the subsequent examination of their multi-cycling characteristics and related heat damage.
2035
treating all these parameters as equal variables will be prohibitively expensive. Thus the current study has focused on the variation of just four selected parameters, namely, composite host material, applied current level, beam length and actuation cycle. There are three major SMAs such as Cu–Zn–Al, Cu–Ni–Al and binary Ni–Ti (nitinol) used for adaptive structural applications [15]. The latter was selected here because of its greater strain recovery capacity, excellent corrosion resistance, stable transformation temperatures with a relatively high electrical resistance and compatibility with cure temperatures of the present host composites. Current binary nitinol wires from Memory-Metalle GmbH had a nickel content of 55.3%. Its transformation temperatures were determined from thermographs generated using a differential scanning calorimetre (DSC). They are 16 °C for Mf, 21 °C for Ms, 47 °C for As, and 55 °C for Af. While a Young’s modulus of 51.8 GPa was obtained from an austenitic stress–strain response performed at 80 °C, a martensitic Young’s modulus of 26.5 GPa was obtained with a yield strength of 268 MPa at 1.96%. Nitinol wires of two different diameters (0.25 mm and 0.51 mm) were evaluated for their strain recovery capability as well as their actuation performance. Although the thinner wire in the construction of smart beams caused the less local distortion to the composite hosts (even when embedded in a unfavourable manner with a very small wire-to-wire spacing), as shown in Fig. 1, the thicker wire of 0.51 mm diameter was selected for its greater potential of recovery force for a constant wire volume fraction [16]. It was understood that the greater energy density via the larger diameter wires could be achieved at the expense of slower heating and cooling rates due to their increased mass and need for thermal transfer. Two different composite host materials, carbon/epoxy and Eglass/epoxy, were evaluated. Although a 32-ply thick quasi-isotropic carbon/epoxy laminate would provide a more useful flexural rigidity for the host beam as well as a larger eccentricity, its flexural rigidity could be much more difficult to overcome. Thus, a 16-ply carbon/epoxy laminate (T700/LTM45-EL) was used as the host with a stacking sequence of (45°/90°/ 45°/0°/SMA/0°/ 45°/ 90°/45°/45°/90°/ 45°/0°/0°/ 45°/90°/45°) with a nominal ply thickness of 0.128 mm. The flexural modulus of the intact host laminate beams was measured to be about 47 GPa. The flexural modulus of the smart beams with the wires being in the martensitic condition was reduced to about 44 GPa. One through-thethickness quarter location was selected to provide a wire eccentricity, although it was very desirable to have the wires embedded as far away from the mid-plane of the beams as possible. Since the wire diameter was about four times greater than the nominal ply thickness, two adjacent plies were oriented to 0° in the longitudinal direction of the beams so that the nitinol wires could partially sink in between them so as to minimise local distortion, as a micrograph in Fig. 2 shows. The local waviness in the micrograph is still quite visible. Nevertheless, the previous experimental investiga-
2. Design considerations The design and manufacture of adaptive composite beams require considerations and selection of a significant number of performance and actuation parameters. These include SMA wire material, wire cross-sectional profile and dimension, prestrain level, volume fraction and through-the-thickness location, host composite material and lay-up, a level of applied current, and beam dimensions for the given performance requirements in terms of bending strain. A comprehensive experimental investigation of
Fig. 1. A micrograph of showing three embedded nitinol wires of 0.25 mm diameter within carbon/epoxy host.
2036
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
Fig. 2. A micrograph of showing two embedded nitinol wires of 0.51 mm diameter within carbon/epoxy host.
tion for the through-the-thickness static mechanical properties of smart beams in [14] confirmed that neither interlaminar shear nor flexural properties suffered substantial degradation. Although the optimum curing temperature of 60 °C and a glass-transition temperature of 78 °C for the matrix were slightly greater than austenitic start and finish temperatures, nitinol wires were tied up on the wire alignment frame during cure so that potential movement through actuation during composite cure could be minimised and thereby their prestrain would not be affected. In addition, it was understood that there was possibility that electricity could be leaked to some extent during actuation because the nitinol wires were not shielded against semi-conductive carbon fibres. An option of shielding the nitinol wires with thermoplastic tubing was considered initially but was not adopted because it could be too intrusive to the composite hosts whose intended nominal thickness was only around 2 mm. The E-glass/epoxy system (SE 84LV/EGL) had a much thicker nominal ply thickness of 0.37 mm. Because of this, the only eight plies were used to construct a symmetric quasi-isotropic lay-up in a stacking sequence of (45°/0°/SMA/ 45°/90°/90°/ 45°/0°/ 45°). In this way, the wire eccentricity of slightly thicker E-glass/ epoxy beams would be same as that of the carbon/epoxy beams. A 0.51 mm diameter wire was partially sunk into one 0° ply at the 0°/ 45° interface. The local distortion associated with the nitinol wire embedment was believed to be less than the case of the carbon/epoxy host. The flexural moduli of the intact host laminate and smart beams with the wires being in the martensitic condition were measured to be about 26 GPa and 23 GPa, respectively. A beam width of 25 mm was selected to allow a maximum number of ten nitinol wires to be embedded side by side with a minimum wire-to-wire spacing of 1.5 mm. This spacing was estimated through an experimental microscopic study [14]. As a result, this resulted in the wire volume fraction of about 4.1%. As this level of wire volume fraction was low, thus an actuation response rate was expected to be relatively slow. However, it could be enhanced
with properly selected levels of actuation power. The level of prestrain in the nitinol wires was measured to be about 5.5% with the intention to provide a maximum energy density. This was selected on the basis of a understanding of the fact that if prestrained beyond 6% the wires could risk significant actuation loss via ageing and damage [4,17]. Three different wire lengths (130 mm, 180 mm and 250 mm) were used to examine the effect of actuation response rates. 3. Specimen manufacture To prepare nitinol wires for prestraining, each piece of nitinol wires was first suspended under a constant load at the side of a 6-m tower before embedment and was trained by cycling it through transformation temperatures via resistive heating for a total of 25 times. The wire was then suspended again on the tower after training and was finally loaded incrementally to a prestain level with dead weights and the corresponding wire elongations being recorded. To achieve a desired plastic prestrain of 5.5%, the wire had to be slightly over-stretched to a strain of about 6% as it contracted slightly when the weights were removed from the wires. An alignment frame device as shown in Fig. 3 was designed and made to ensure that during composite cure the prestrained wires were kept not only longitudinally straight with the desired wireto-wire spacing but were also at the desired through-the-thickness location. It consisted of an aluminium base place; a pair of through-the-thickness positioning blocks; a pair of wire tension tightening blocks; and two arrays of wire fixers (one array metal ferrules and one array of nuts). Each through-the-thickness positioning block had the same number of orifices as the nitinol wires and the diameter of each orifice was fractionally greater than the wire diameter so that the wires were threaded through easily without being damaged. To prepare the wire for a single specimen, one end of the wire was threaded though the two positioning blocks, which were
Fig. 3. SMA wire alignment device for manufacturing adaptive composite beams.
2037
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
4. Experimental set-up and procedure Each of the adaptive composite beams was fixed at one end in the cantilever set-up. The free end deflection of each beam was measured by a high-precision LVDT with an accuracy of ±0.1 mm. The actual contact location of the bearing ball of LVDT was about 2 mm away from the physical end of each beam. Alternatively, a transverse bending force could also be measured if a load cell was placed in the same position as the LVDT. High-temperature strain gauges (SGs) were bonded on both surfaces of some selected beams. The overall experimental set-up is shown in Fig. 4. In addition to the direct temperature measurement of embedded wires, temperatures of both the beam surface and the protruding part of the nitinol wires of some beams were also measured with thermocouples. To actuate each beam, electric current of selected level
was applied to the nitinol wire via a programmable low-voltage power supply of model Xantrex 300-3.5A. The duration and the level of current applied to the present adaptive beams were determined through the establishment of a deflection plateau on the deflection–time curves, as seen in Figs 5, 9 and 10. The power was disconnected either after the estimated deflection plateau was reached or before in order to avoid heat damage, depending on whether the beams were intended for single-cycling or multicycling. The adaptive beams were either left to cool at room temperature (RT). 5. Results and discussion 5.1. Smart E-glass/epoxy beams E-glass/epoxy beams with a relatively low flexural rigidity were expected to show a substantial level of deflection. As they had three different lengths, their respective deflections were expected to vary, depending on specific beam length and/or current level. To this end, several applied current levels were tried, varying from 0.5 A to 2 A. A deflection summary of all tests on smart E-glass/ epoxy beams of various lengths is presented in Table 1. Fig. 5 shows a typical end deflection–time curve from a 180 mm long
30
Heating
Cooling
25
End deflection (mm)
slid along the wires such that the wire tension was taut. Then the wire was clamped through the tightening blocks and end nuts. For the carbon/epoxy host, a total of 16 plies for each adaptive beam were laid up in two separate sub-laminate stacks. The lower stack consisted of 12 plies whereas the upper stack consisted of four plies. To assemble a whole laminate stack, all the usual vacuum bagging materials were prepared. After sliding the thicker lower stack through below the wires, the upper stack was placed on the top of the wire along with all necessary vacuum bagging materials. The entire laminate assembly was cured in an autoclave, following the manufacturers recommended cycle of 60 °C at 90 psi for 18 h. For the E-glass/epoxy host, the assembly process was the same and the manufacturers recommended cure cycle was 90 °C at 90 psi for 6 h. For each type of the hosts, three beams each with a different length were manufactured, and two or three specimens were duplicated for each length. In a few selected beams, K-type thermocouples were also bonded onto the outermost wire with high temperature cement within the host so that the temperature of the embedded wire could directly be measured. Each one was oriented to the direction perpendicular to the nitinol wire at the central section so that its potential separation from the wire was minimised. After cure, each specimen was removed from the alignment device. The wires of the specimen were cut in such a way that they were about 10 mm longer than the smart beam at both ends. After having epoxy cleaned up, the wires were connected by metal ferrules to form an electric circuit (see the right-hand side of the photograph in Fig. 7).
20
15
10
5
0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 5. Deflection–time curve from a 180 mm adaptive E-glass/epoxy beam at RT of 22 °C.
Fig. 4. An experimental set-up for actuation of an adaptive beam embedded with nitinol wires.
2038
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
Table 1 Deflections of E-glass/epoxy beams embedded with 0.51 mm nitinol wires. Current level (A)
1.0 1.5 2.0
Free end deflection of three types of beams 130 mm
180 mm
250 mm
2.92 ± 0.08 8.36 ± 0.54 13.21
– 15.38 ± 0.77 20.81 ± 1.36
– 28.48 ± 1.36 41.93 ± 2.02
E-glass/epoxy beam with an applied current of 2 A. The reverse phase transformation seemed more or less complete after 5 min when the end deflection reached an asymptote of about 23 mm. As the present thick wires were well bonded with the host, the lengthened response time was expected, to some extent. Also it seemed obvious that a natural cooling could take much longer for the beam to return to its starting position. Fig. 6 shows a photograph of a 130 mm long E-glass/epoxy beam with its end deflection reaching about 8 mm, when activated with an applied current of 1.5 A. It was found from smart E-glass/epoxy beams that generally the longer the beam was, the greater end deflection it could be obtained for the given applied current level with the maximum deflection of over 42 mm being achieved from one 250 mm long beam activated at 2 A. The shortest duration for an E-glass/epoxy beam to achieve an asymptote was around 3 min from one 130 mm long beam with the maximum deflection of about 13 mm. In particular, it was encouraging to observe that in the first 11 mm deflection out of the total 13 mm in this test the response of the beam was more or less linear with an approximate rate of 6 mm per minute.
From the above single cycle experimental exercises with beam length and applied current level, it seemed that the applied current level of 2 A actuated the beams with the relatively faster responsive rates. In particular, achieving the maximum deflections within the relatively short times could mean that the less heat transfer was needed for cooling to be completed. An applied current of 2 A was thus used for multi-cycling other E-glass/epoxy beams with the intention to achieve a maximum end deflection in each cycle similar to those single cycle exercises. However, a wire-composite debonding could take place as shown in Fig. 7 in an E-glass/ epoxy beam after just 15 cycles in some cases. There translucent Eglass fibres allowed the debonding damage to be clearly seen with the front of the debonded area being semi-circular. Although the higher applied current level provided the higher wire temperature, those wires around the middle region of beams could have had slightly higher temperatures because they were less able to dissipate the heat. At this current level, a 250 mm E-glass/epoxy beam fractured at the 5th cycle due to combined thermal and mechanical stresses, as shown in Fig. 8. As a result, subsequent multi-cycling actuations were conducted at an applied current of 1.5 A and with the power being cut off after 5 min to avoid damage. 5.2. Smart carbon/epoxy beams A strong focus of the investigation was placed on the carbon/ epoxy host, as this has been the primary composite system used in the aerospace industry, though the nitinol wires were not shielded to prevent potential contact with semi-conductive carbon fibres. Fig. 9 shows a typical end deflection–time curve from a 130 mm carbon/epoxy beam (A) with a current of 3.5 A for the
Fig. 6. A photograph showing actuation of a 130 mm E-glass/epoxy beam at RT.
Fig. 7. A photograph showing the front of the debonded region of an E-glass/epoxy beam due to multiple-cycling.
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
2039
Fig. 8. A photograph showing a failed E-glass/epoxy beam due to multiple-cycling.
1.2
Heating
Cooling
End deflection (mm)
1.0 0.8 0.6 0.4 0.2 0.0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 9. Deflection–time curve from a 130 mm adaptive carbon/epoxy beam at RT of 22 °C.
1.2
Heating
Cooling
End deflection (mm)
1.0
Ambient temperature
0.8
0
of 18 C
shielded carbon fibres could have got in contact with the nitinol wires so that a large part of the supplied current was lost. Nevertheless, when the power was cut off, the nitinol wires cooled exponentially through conduction and convection. The flexural rigidity of the adaptive beam acted as a conventional spring to return the beam to its undeformed position. However, the residual deflection of about 23% maximum end deflection at the end of cooling is visible in Fig. 9, though some tests had much smaller residual deflections (see Fig. 16). This noticeable non-zero deflection could be attributed to the difference between Mf of 16 °C and ambient (or room) temperature of about 20 °C so that the forward transformation was likely to be incomplete. In particular, the actual Mf of the prestrained wires could be less than that tested with zero prestrain [18]. To verify the latter, a series of tests on the same beam were repeated in an open freezer with ambient temperature of being close to 18 °C. Fig. 10 shows a comparison of two deflection–time curves, in which the residual deflection of the test with the lower ambient temperature diminished to the only 5% maximum end deflection. This could be attributed to the increased forward transformation during the cooling period. The maximum end deflection somewhat was also reduced by the variation of ambient temperatures. Nevertheless, when both curves were normalised with reference to the respective peak deflections, the early observation was verified in Fig. 11. Indeed, the beam tested in the freezer in Fig. 10 took a little longer time to reach the same deflection in
0.6 0.4
Ambient temperature
0.2
0
of -18 C 0.0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 10. Deflection–time curves from a 130 mm carbon/epoxy beam actuated at two different ambient temperatures.
duration of 7 min. The initial linear region lasted only up to about 0.35 mm in the first 50 s. The rest of the curve exhibits a nonlinear trend and levelled off with a plateau which corresponded to the maximum end deflection of 0.89 mm. Although this level of deflection seemed very small, it could indicate that the bending moment induced by the recovery force was balanced out by the flexural rigidity of the adaptive beam. It was equally possible that non-
Fig. 11. Normalised deflection–time curves of a 130 mm carbon/epoxy beam tested at two different ambient temperatures.
2040
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
the reverse transformation and a lesser time to cool down. To establish repeatability of the above actuation performance and consistency of the fabrication process, the entire experimentation was repeated with another 130 mm carbon/epoxy beam (B) fabricated and tested in the same way. In addition, a thermocouple was also embedded in this beam. This second adaptive beam B achieved the maximum end deflection of 0.88 mm at RT with the end deflection–time curves of both beams shown in Fig. 12. Clearly, such repeatable actuation performance from these adaptive beams confirmed once again that the present fabrication process was very adequate and reliable. Fig. 13 shows the temperature–time curve of the 130 mm carbon/epoxy beam B, in which both wire and beam surface temperatures were measured by embedded and surface-mounted thermocouples, respectively. It can be seen that the wire temperature rose almost linearly to about 195 °C in 20 s and reached the maximum temperature of about 215 °C in 50 s. It remained at that level for the rest of the heating period. Clearly, the plateau temperature was nearly four times Af of 55 °C. It implies that much higher temperature and overheating was necessary to complete the reverse transformation for prestrained wires [18] and for larger recovery stress [19,20]. The cooling response was very similar to the heating one, i.e. the wire temperature was back to the initial temperature in a similar time scale. Interestingly, the temperature on the compressive beam surface arose steadily to about 22 °C after 50 s and reached 40 °C only when cooling started. As a nominal distance between the embedded wires and the closer surface
was only about 0.25 mm, this seems to indicate that these carbon/epoxy beams were very efficient thermal insulators with the limited ability to dissipate interior heat. If such heating duration was substantial and cyclic, heat damage of epoxy could indeed become an issue. When the above temperature response was plotted against the corresponding deflection as in Fig. 14, it became clear that the deflection started only when the stress-free Af was passed and that major part of deflection took place when the plateau temperature of 215 °C was reached. In the present actuation evaluation of the adaptive beams, the examination of the effectiveness of eccentric axial load-bending transformation is extremely crucial. As the load transfer from contraction of the embedded nitinol wires to the host bending was via interfacial shear, thus the most effective region of the load transfer would be at the free end of the adaptive beams since the other end was always clamped. This interfacial shear was likely to decay towards the middle of the beam. Fig. 15 shows strain responses of two back-to-back SGs 26 mm from the clamping end. Interestingly, the strains from both SGs were of compressive nature. The peak values are 230 le from the surface remote to the wires and 455 le from the surface closer to the wires. On the basis of the superposition principle in elementary beam theory, the bending strain at the same location could be only 113 le whereas the compressive axial strain was about 343 le, which was three times greater. This seems to suggest that not all the recovery strain generated by the wire contraction was transferred to the induction of bending strain. 1.2 1.1
1.2
Heating
1.1
1.0
Cooling End deflection (mm)
End deflection (mm)
1.0 0.9 0.8 0.7 0.6
Specimen A
0.5 0.4
Specimen B
0.9 0.8 0.7 0.6 0.5 0.4 0.3
0.3
Stress-free M f
0.2
0.2
Stress-free A f
0.1
0.1
0.0
0.0 0
100
200
300
400
500
600
700
800
900
0
1000
25
50
Fig. 12. Deflection–time curves from two 130 mm carbon/epoxy beams at room temperature.
250
Heating
125
150
175
200
225
250
Fig. 14. Deflection–temperature curve from a 130 mm adaptive carbon/epoxy beam B at RT.
Cooling 100
200
50
175 150 125
-50
100
200
50
-200
200
300
500
600
700
800
900
to the nitinol wires
-300
Straingauge close to
-400
100
400
Strain gauge remote
-250 -350
25 0
300
-150
Microstrain
75
0
-100
Temperature of the beam surface closer to the wires
100
Time (second)
0
Embedded wire temperature
0
Temperature ( C)
100
0
225
0
75
Wire temperature ( C)
Time (second)
400
500
600
700
800
900
1000
Time (second)
-500 -550
Fig. 13. Temperature–time curves from a 130 mm adaptive carbon/epoxy beam B at RT.
the nitinol wires
-450
Heating
Cooling
Fig. 15. Strain–time curves from a 130 mm carbon/epoxy beam.
1000
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041 2.0
Heating
1.6
End deflection (mm)
tion capability derived from the single-cycle actuation exercises was not generally suited to that for multi-cycling actuation and could result in premature failure of multi-cycled smart beams.
Cooling
1.8
2041
250-mm beam
1.4
Acknowledgement
1.2
180-mm beam
1.0
The first author acknowledges the execution of some tests by Mr. A.R. Giles.
0.8
130-mm beam
0.6
References
0.4 0.2 0.0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 16. Deflection–time curves from adaptive carbon/epoxy beams of three different lengths at room temperature.
Beams of two different additional lengths were also tested and their deflection results as included in Fig. 16 show that both the initial linear regions and the maximum end deflections seemed to be proportional to the increase in wire length, as an increase in beam length increased the electric resistance and thermal capacitance of the wires as also reported in [21]. 6. Conclusions Smart adaptive composite beams embedded with prestrained nitinol wire actuators were designed and fabricated in an autoclave with a purpose-made alignment device. The developed fabrication procedure overcame a number of technical challenges such that prestrained wires were successfully embedded eccentrically in parallel in the length direction of composite beams with three different lengths. Both single-cycle and multi-cycle thermomechanical bending actuations of these adaptive beams in the cantilever set-up were characterised experimentally by applying various levels of electric current to the nitinol wires. Such characterisation showed that the present fabrication process was repeatable and reliable. While the significant end deflections of up to 41 mm were easily achieved from the smart E-glass/epoxy beams, the limited end deflections were observed from the smart carbon/epoxy beams due primarily to our inability to find a workable solution of insulating the nitinol wires. Moreover, it seemed necessary to overheat the prestrained wires to much higher temperatures beyond the complete reverse transformation in order to generate recovery stress. The longer beams with greater actuation rates took less time to reach the same level of deflection. It was found that the actua-
[1] Birman V. Review of mechanics of shape memory alloy structures. Appl Mech Rev 1997;50:629–45. [2] Wei ZG, Sandstrom R, Miyazaki S. Shape memory materials and hybrid composites for smart system. J Mater Sci 1998;33:3763–83. [3] Baz A. Shape control of composites. In: Kelly A, Zweben C, editors. Comprehensive composite materials. London: Elsevier Science; 2000. [4] Chopra I. Review of state of art of smart structures and integrated systems. AIAA J 2002;40:2145–87. [5] Turner TL, Lach CL, Cano RJ. Fabrication and characterization of SMA hybrid composites. Proc SPIE Conf 2001:4333–60. [6] Moore CL, Bruck HA. A fundamental investigation into large strain recovery of one-way shape memory alloy wires embedded in flexible polyurethanes. Smart Mater Struct 2002;11:130–9. [7] Davis BA. Investigation of the thermomechanical response of shape memory alloy hybrid composite beams. NASA/CR-2005-213929. [8] Turner TL, Buehrle RD, Cano RJ, Fleming GA. Modeling, fabrication and testing of a SMA hybrid composite jet engine Chevron concept. J Intell Mater Syst Struct 2006;17:483–97. [9] Yang SM, Roh JH, Han JH, Lee I. Experimental studies on active shape control of composite structures using SMA actuators. J Intell Mater Syst Struct 2006;17:767–77. [10] Hebda DA, Whitlock ME, Ditman JB, White SR. Manufacturing of adaptive graphite/epoxy structures with embedded Nitinol wires. J Intell Mater Syst Struct 1995;6:220–8. [11] de Blonk BJ, Lagoudas DC. Actuation of elastomeric rods with embedded twoway shape memory alloy actuators. Smart Mater Struct 1998;7:771–83. [12] Doran CJ. The effect of prestrain on the actuation performance of embedded shape memory alloy wires. In: Proc of 2nd Euro conf on smart struct mater, Glasgow; 1994. [13] Friend CM, Morgan NB, Wise VD. Some durability issues associated with the use of shape-memory alloy ‘smart’ composites. Proc Mater Res Soc Symp 1995:513–8. [14] Zhou G, Sim LM, Brewster P, Giles AR. Through-the-thickness mechanical properties of quasi-isotropic carbon/epoxy laminates. Composites Part A 2004;35:797–815. [15] Huang W. On the selection of shape memory alloys for actuators. Mater Design 2002;23:11–9. [16] Epps J, Chandra R. Shape memory alloy actuation for active tuning of composite beams. Smart Mater Struct 1997;6:251–64. [17] Brinson LC, Huang MS, Boller C, Brand W. Analysis of controlled beam deflections using SMA wires. J Intell Mater Syst Struct 1997;8:12–25. [18] Stalman R, Tsoi K, Schrooten J. The transformation behaviour of shape memory wires embedded in a composite matrix. Proc SPIE Conf 2000;4073:88–96. [19] Brinson LC, Bekker A, Hwang S. Temperature induced deformation in shape memory alloys. Proc SPIE Conf 1994;2427:234–44. [20] Elspass WJ, Kunzmann J. Design, manufacturing and verification of laminates with embedded SMA actuators. Proc SPIE Conf 1996;2717:320–9. [21] Baz A, Inman K, McRoy J. The dynamic and thermal characteristics of shape memory actuators. Proc SPIE Conf 1994;2190:436–53.
Contents lists available at ScienceDirect
Composites Science and Technology journal homepage: www.elsevier.com/locate/compscitech
Design, manufacture and evaluation of bending behaviour of composite beams embedded with SMA wires Gang Zhou a,*, Peter Lloyd b a b
Department of Aeronautical and Automotive Engineering, Loughborough University, Loughborough, Leicestershire LE11 3TU, UK DSTL, The Ministry of Defence, Farnborough, Hants GU14 0LX, UK
a r t i c l e
i n f o
Article history: Received 15 July 2008 Received in revised form 15 January 2009 Accepted 18 January 2009 Available online 25 January 2009 Keywords: A. Smart materials B. Stress/strain B. Thermomechanical properties C. Deformation Shape memory effect
a b s t r a c t An experimental study has been conducted to design and fabricate smart composite beams embedded with prestrained nitinol wire actuators. The developed fabrication process allowed both quasi-isotropic E-glass/epoxy and carbon/epoxy hosts to be eccentrically embedded with 10 parallel prestrained wires with a purpose-made alignment device and cured successfully in an autoclave. Smart composite beams of three different lengths were made for each type of host. Both single-cycle and multi-cycle thermomechanical bending actuations of these beams in the cantilever set-up were characterised experimentally by applying various levels of electric current to the nitinol wires. The performance characteristics showed that the present fabrication process was repeatable and reliable. While the end deflections of up to 41 mm were easily achieved from smart E-glass/epoxy beams, the limited end deflections were observed from the smart carbon/epoxy beams due primarily to our inability to insulate the nitinol wires. Moreover, it seemed necessary to overheat the prestrained wires to much higher temperatures beyond the complete reverse transformation in order to generate recovery stress. The longer beams showed greater actuation rates and took less time to reach the same level of deflection. It was found that the actuation capability derived from single-cycle actuation exercises was not suited to multi-cycling actuations and could result in premature failure of multi-cycled smart beams. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction Shape memory alloys (SMAs) such as nitinol have the ability to exhibit diffusionless solid-state phase transformations, which are induced either by mechanical stress or by temperature. Commercially, they are available in wire or ribbon form. The two solid-state crystalline phases involved are room temperature martensite and high temperature parent phase austenite. Their unique transformations from austenite to martensite are characterised by four critical temperatures, i.e. martensitic start temperature Ms, martensitic finish temperature Mf, austenitic start temperature As and austenitic finish temperature Af. One type of transformation, called pseudoelasticity, superelasticity, or stress-induced martensite (SIM), occurs when SMAs in the isothermal austenitic condition are mechanically stressed into the plastic region so that their plastic strain is completely recovered on the removal of the stress. The other type of transformation, called the shape memory effect (SME), occurs when plastically deformed SMAs in the martensite state recovers their original shape (i.e. plastic strain), if heated electrically above their Af. If the recovery of the plastic strain is constrained, a large recovery stress will be generated. Since their * Corresponding author. Tel.: +44 01509 227 210; fax: +44 01509 227 275. E-mail address: [email protected] (G. Zhou). 0266-3538/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.compscitech.2009.01.017
discovery in the early 60s, a great deal of effort has been devoted to characterising SMAs, exploiting these two thermomechanical phenomena and developing new applications in aerospace, civil, mechanical and medical industries as discussed in [1–9]. In particular, as those transformation temperatures for any given SMAs can be tuned to suit a specific application via a change of the material composition and/or heat treatment, they can be developed into smart actuators through extensive exploitations of the recovery stress for smart adaptive structure technology. In addition to the aforementioned one-way SME, of current interest, SMAs could be trained to exhibit the two-way SME as reported in [10,11], but with a much limited recovery strain. To develop smart adaptive structural components by controlling their shapes in bending, SMA actuators in general need to be integrated with the host components such as composite structures, though some non-embedding approaches have also been investigated [1–4,9]. Important performance parameters for such smart components include those associated with SMA actuators such as maximum strain, prestrain, wire diameter, energy density ratio, stiffness, bandwidth (including cooling rate) and actuation capability; and those associated with the host such as actuator volume fraction, actuator through-the-thickness location, host flexural rigidity, actuator–host interfacial strength and actuation cycle. For the smart components to be adaptive, the respective contribu-
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
tions of the aforementioned parameters to their functional performance must be synergistic, which can be a significant challenge. This is because the functional attributes of their performance contributions sometimes lead to significant conflicting consequences, which compromise the performance of the components, as well as undesirable ‘side effects’. For example, an adaptive beam embedded with prestrained one-way SMA wires was expected nominally to show substantial deflection with a reasonable actuation rate, once energised through electric heating. Intuitively, it seemed desirable for the host composite beam to (a) have a large amount of highly prestrained wires embedded with a sufficiently large eccentricity to deliver sufficient bending moment and then substantial deflection; (b) have a relatively moderate flexural rigidity against which the beam bending induced by the not-so-high level of bending moment would have to overcome; and (c) be actuated by a high level of electric current to achieve desired actuation responsive rate. However, these attributes led immediately to several difficulties as they were conflicting requirements. Although nitinol wires had a recovery strain of up to 8%, the high level of prestraining could lead to their rapid ageing during multi-cycle actuations [12]. The thickness of the composite host remained the most difficult parameter to reconcile. One the one hand, a relatively large thickness offered a greater eccentricity and a relatively large flexural rigidity. The latter was required not only to return the actuated beam to its original position but also to prestrain the embedded wire actuators again for the next cycle. On the other hand, the greater flexural rigidity of the beam could become a greater obstacle for the host to overcome to achieve the desired bending. Moreover, as phase transformations were hysteretic and could thus release latent heat, a high level of applied current, desirable and necessary for a high actuation response rate, could result in heat damage in the composite host [12,13], which could in turn affect the wire actuation capability for subsequent cycles. In addition, embedding too many SMA wires in the host could also adversely affect the long-term through-thethickness mechanical properties of the host structures [14]. Although the literature abounds with individual endeavours of using SMAs for the shape control of smart adaptive structures as discussed in [1–9], a synergistic set of the performance attributes from the aforementioned parameters seemed clearly application specific involving inevitable compromises. In particular, the experimental thermomechanical behaviour of embedded SMA actuators, underpinning and justifying those compromises, has not been understood. To this end, it is of paramount importance to develop a good understanding of actuation characteristics of SMA-based adaptive composite beams in terms of performance bounds of the aforementioned important parameters. This experimental study aims on the examination of actuation characteristics of Eglass/epoxy and carbon/epoxy laminate beams embedded with nitinol wire actuators. Its focus will be on the manufacturability and actuation repeatability of adaptive composite beams and the effects of beam length and applied current level on their actuation characteristics. Information generated and experience gained will be channelled into the subsequent examination of their multi-cycling characteristics and related heat damage.
2035
treating all these parameters as equal variables will be prohibitively expensive. Thus the current study has focused on the variation of just four selected parameters, namely, composite host material, applied current level, beam length and actuation cycle. There are three major SMAs such as Cu–Zn–Al, Cu–Ni–Al and binary Ni–Ti (nitinol) used for adaptive structural applications [15]. The latter was selected here because of its greater strain recovery capacity, excellent corrosion resistance, stable transformation temperatures with a relatively high electrical resistance and compatibility with cure temperatures of the present host composites. Current binary nitinol wires from Memory-Metalle GmbH had a nickel content of 55.3%. Its transformation temperatures were determined from thermographs generated using a differential scanning calorimetre (DSC). They are 16 °C for Mf, 21 °C for Ms, 47 °C for As, and 55 °C for Af. While a Young’s modulus of 51.8 GPa was obtained from an austenitic stress–strain response performed at 80 °C, a martensitic Young’s modulus of 26.5 GPa was obtained with a yield strength of 268 MPa at 1.96%. Nitinol wires of two different diameters (0.25 mm and 0.51 mm) were evaluated for their strain recovery capability as well as their actuation performance. Although the thinner wire in the construction of smart beams caused the less local distortion to the composite hosts (even when embedded in a unfavourable manner with a very small wire-to-wire spacing), as shown in Fig. 1, the thicker wire of 0.51 mm diameter was selected for its greater potential of recovery force for a constant wire volume fraction [16]. It was understood that the greater energy density via the larger diameter wires could be achieved at the expense of slower heating and cooling rates due to their increased mass and need for thermal transfer. Two different composite host materials, carbon/epoxy and Eglass/epoxy, were evaluated. Although a 32-ply thick quasi-isotropic carbon/epoxy laminate would provide a more useful flexural rigidity for the host beam as well as a larger eccentricity, its flexural rigidity could be much more difficult to overcome. Thus, a 16-ply carbon/epoxy laminate (T700/LTM45-EL) was used as the host with a stacking sequence of (45°/90°/ 45°/0°/SMA/0°/ 45°/ 90°/45°/45°/90°/ 45°/0°/0°/ 45°/90°/45°) with a nominal ply thickness of 0.128 mm. The flexural modulus of the intact host laminate beams was measured to be about 47 GPa. The flexural modulus of the smart beams with the wires being in the martensitic condition was reduced to about 44 GPa. One through-thethickness quarter location was selected to provide a wire eccentricity, although it was very desirable to have the wires embedded as far away from the mid-plane of the beams as possible. Since the wire diameter was about four times greater than the nominal ply thickness, two adjacent plies were oriented to 0° in the longitudinal direction of the beams so that the nitinol wires could partially sink in between them so as to minimise local distortion, as a micrograph in Fig. 2 shows. The local waviness in the micrograph is still quite visible. Nevertheless, the previous experimental investiga-
2. Design considerations The design and manufacture of adaptive composite beams require considerations and selection of a significant number of performance and actuation parameters. These include SMA wire material, wire cross-sectional profile and dimension, prestrain level, volume fraction and through-the-thickness location, host composite material and lay-up, a level of applied current, and beam dimensions for the given performance requirements in terms of bending strain. A comprehensive experimental investigation of
Fig. 1. A micrograph of showing three embedded nitinol wires of 0.25 mm diameter within carbon/epoxy host.
2036
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
Fig. 2. A micrograph of showing two embedded nitinol wires of 0.51 mm diameter within carbon/epoxy host.
tion for the through-the-thickness static mechanical properties of smart beams in [14] confirmed that neither interlaminar shear nor flexural properties suffered substantial degradation. Although the optimum curing temperature of 60 °C and a glass-transition temperature of 78 °C for the matrix were slightly greater than austenitic start and finish temperatures, nitinol wires were tied up on the wire alignment frame during cure so that potential movement through actuation during composite cure could be minimised and thereby their prestrain would not be affected. In addition, it was understood that there was possibility that electricity could be leaked to some extent during actuation because the nitinol wires were not shielded against semi-conductive carbon fibres. An option of shielding the nitinol wires with thermoplastic tubing was considered initially but was not adopted because it could be too intrusive to the composite hosts whose intended nominal thickness was only around 2 mm. The E-glass/epoxy system (SE 84LV/EGL) had a much thicker nominal ply thickness of 0.37 mm. Because of this, the only eight plies were used to construct a symmetric quasi-isotropic lay-up in a stacking sequence of (45°/0°/SMA/ 45°/90°/90°/ 45°/0°/ 45°). In this way, the wire eccentricity of slightly thicker E-glass/ epoxy beams would be same as that of the carbon/epoxy beams. A 0.51 mm diameter wire was partially sunk into one 0° ply at the 0°/ 45° interface. The local distortion associated with the nitinol wire embedment was believed to be less than the case of the carbon/epoxy host. The flexural moduli of the intact host laminate and smart beams with the wires being in the martensitic condition were measured to be about 26 GPa and 23 GPa, respectively. A beam width of 25 mm was selected to allow a maximum number of ten nitinol wires to be embedded side by side with a minimum wire-to-wire spacing of 1.5 mm. This spacing was estimated through an experimental microscopic study [14]. As a result, this resulted in the wire volume fraction of about 4.1%. As this level of wire volume fraction was low, thus an actuation response rate was expected to be relatively slow. However, it could be enhanced
with properly selected levels of actuation power. The level of prestrain in the nitinol wires was measured to be about 5.5% with the intention to provide a maximum energy density. This was selected on the basis of a understanding of the fact that if prestrained beyond 6% the wires could risk significant actuation loss via ageing and damage [4,17]. Three different wire lengths (130 mm, 180 mm and 250 mm) were used to examine the effect of actuation response rates. 3. Specimen manufacture To prepare nitinol wires for prestraining, each piece of nitinol wires was first suspended under a constant load at the side of a 6-m tower before embedment and was trained by cycling it through transformation temperatures via resistive heating for a total of 25 times. The wire was then suspended again on the tower after training and was finally loaded incrementally to a prestain level with dead weights and the corresponding wire elongations being recorded. To achieve a desired plastic prestrain of 5.5%, the wire had to be slightly over-stretched to a strain of about 6% as it contracted slightly when the weights were removed from the wires. An alignment frame device as shown in Fig. 3 was designed and made to ensure that during composite cure the prestrained wires were kept not only longitudinally straight with the desired wireto-wire spacing but were also at the desired through-the-thickness location. It consisted of an aluminium base place; a pair of through-the-thickness positioning blocks; a pair of wire tension tightening blocks; and two arrays of wire fixers (one array metal ferrules and one array of nuts). Each through-the-thickness positioning block had the same number of orifices as the nitinol wires and the diameter of each orifice was fractionally greater than the wire diameter so that the wires were threaded through easily without being damaged. To prepare the wire for a single specimen, one end of the wire was threaded though the two positioning blocks, which were
Fig. 3. SMA wire alignment device for manufacturing adaptive composite beams.
2037
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
4. Experimental set-up and procedure Each of the adaptive composite beams was fixed at one end in the cantilever set-up. The free end deflection of each beam was measured by a high-precision LVDT with an accuracy of ±0.1 mm. The actual contact location of the bearing ball of LVDT was about 2 mm away from the physical end of each beam. Alternatively, a transverse bending force could also be measured if a load cell was placed in the same position as the LVDT. High-temperature strain gauges (SGs) were bonded on both surfaces of some selected beams. The overall experimental set-up is shown in Fig. 4. In addition to the direct temperature measurement of embedded wires, temperatures of both the beam surface and the protruding part of the nitinol wires of some beams were also measured with thermocouples. To actuate each beam, electric current of selected level
was applied to the nitinol wire via a programmable low-voltage power supply of model Xantrex 300-3.5A. The duration and the level of current applied to the present adaptive beams were determined through the establishment of a deflection plateau on the deflection–time curves, as seen in Figs 5, 9 and 10. The power was disconnected either after the estimated deflection plateau was reached or before in order to avoid heat damage, depending on whether the beams were intended for single-cycling or multicycling. The adaptive beams were either left to cool at room temperature (RT). 5. Results and discussion 5.1. Smart E-glass/epoxy beams E-glass/epoxy beams with a relatively low flexural rigidity were expected to show a substantial level of deflection. As they had three different lengths, their respective deflections were expected to vary, depending on specific beam length and/or current level. To this end, several applied current levels were tried, varying from 0.5 A to 2 A. A deflection summary of all tests on smart E-glass/ epoxy beams of various lengths is presented in Table 1. Fig. 5 shows a typical end deflection–time curve from a 180 mm long
30
Heating
Cooling
25
End deflection (mm)
slid along the wires such that the wire tension was taut. Then the wire was clamped through the tightening blocks and end nuts. For the carbon/epoxy host, a total of 16 plies for each adaptive beam were laid up in two separate sub-laminate stacks. The lower stack consisted of 12 plies whereas the upper stack consisted of four plies. To assemble a whole laminate stack, all the usual vacuum bagging materials were prepared. After sliding the thicker lower stack through below the wires, the upper stack was placed on the top of the wire along with all necessary vacuum bagging materials. The entire laminate assembly was cured in an autoclave, following the manufacturers recommended cycle of 60 °C at 90 psi for 18 h. For the E-glass/epoxy host, the assembly process was the same and the manufacturers recommended cure cycle was 90 °C at 90 psi for 6 h. For each type of the hosts, three beams each with a different length were manufactured, and two or three specimens were duplicated for each length. In a few selected beams, K-type thermocouples were also bonded onto the outermost wire with high temperature cement within the host so that the temperature of the embedded wire could directly be measured. Each one was oriented to the direction perpendicular to the nitinol wire at the central section so that its potential separation from the wire was minimised. After cure, each specimen was removed from the alignment device. The wires of the specimen were cut in such a way that they were about 10 mm longer than the smart beam at both ends. After having epoxy cleaned up, the wires were connected by metal ferrules to form an electric circuit (see the right-hand side of the photograph in Fig. 7).
20
15
10
5
0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 5. Deflection–time curve from a 180 mm adaptive E-glass/epoxy beam at RT of 22 °C.
Fig. 4. An experimental set-up for actuation of an adaptive beam embedded with nitinol wires.
2038
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
Table 1 Deflections of E-glass/epoxy beams embedded with 0.51 mm nitinol wires. Current level (A)
1.0 1.5 2.0
Free end deflection of three types of beams 130 mm
180 mm
250 mm
2.92 ± 0.08 8.36 ± 0.54 13.21
– 15.38 ± 0.77 20.81 ± 1.36
– 28.48 ± 1.36 41.93 ± 2.02
E-glass/epoxy beam with an applied current of 2 A. The reverse phase transformation seemed more or less complete after 5 min when the end deflection reached an asymptote of about 23 mm. As the present thick wires were well bonded with the host, the lengthened response time was expected, to some extent. Also it seemed obvious that a natural cooling could take much longer for the beam to return to its starting position. Fig. 6 shows a photograph of a 130 mm long E-glass/epoxy beam with its end deflection reaching about 8 mm, when activated with an applied current of 1.5 A. It was found from smart E-glass/epoxy beams that generally the longer the beam was, the greater end deflection it could be obtained for the given applied current level with the maximum deflection of over 42 mm being achieved from one 250 mm long beam activated at 2 A. The shortest duration for an E-glass/epoxy beam to achieve an asymptote was around 3 min from one 130 mm long beam with the maximum deflection of about 13 mm. In particular, it was encouraging to observe that in the first 11 mm deflection out of the total 13 mm in this test the response of the beam was more or less linear with an approximate rate of 6 mm per minute.
From the above single cycle experimental exercises with beam length and applied current level, it seemed that the applied current level of 2 A actuated the beams with the relatively faster responsive rates. In particular, achieving the maximum deflections within the relatively short times could mean that the less heat transfer was needed for cooling to be completed. An applied current of 2 A was thus used for multi-cycling other E-glass/epoxy beams with the intention to achieve a maximum end deflection in each cycle similar to those single cycle exercises. However, a wire-composite debonding could take place as shown in Fig. 7 in an E-glass/ epoxy beam after just 15 cycles in some cases. There translucent Eglass fibres allowed the debonding damage to be clearly seen with the front of the debonded area being semi-circular. Although the higher applied current level provided the higher wire temperature, those wires around the middle region of beams could have had slightly higher temperatures because they were less able to dissipate the heat. At this current level, a 250 mm E-glass/epoxy beam fractured at the 5th cycle due to combined thermal and mechanical stresses, as shown in Fig. 8. As a result, subsequent multi-cycling actuations were conducted at an applied current of 1.5 A and with the power being cut off after 5 min to avoid damage. 5.2. Smart carbon/epoxy beams A strong focus of the investigation was placed on the carbon/ epoxy host, as this has been the primary composite system used in the aerospace industry, though the nitinol wires were not shielded to prevent potential contact with semi-conductive carbon fibres. Fig. 9 shows a typical end deflection–time curve from a 130 mm carbon/epoxy beam (A) with a current of 3.5 A for the
Fig. 6. A photograph showing actuation of a 130 mm E-glass/epoxy beam at RT.
Fig. 7. A photograph showing the front of the debonded region of an E-glass/epoxy beam due to multiple-cycling.
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
2039
Fig. 8. A photograph showing a failed E-glass/epoxy beam due to multiple-cycling.
1.2
Heating
Cooling
End deflection (mm)
1.0 0.8 0.6 0.4 0.2 0.0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 9. Deflection–time curve from a 130 mm adaptive carbon/epoxy beam at RT of 22 °C.
1.2
Heating
Cooling
End deflection (mm)
1.0
Ambient temperature
0.8
0
of 18 C
shielded carbon fibres could have got in contact with the nitinol wires so that a large part of the supplied current was lost. Nevertheless, when the power was cut off, the nitinol wires cooled exponentially through conduction and convection. The flexural rigidity of the adaptive beam acted as a conventional spring to return the beam to its undeformed position. However, the residual deflection of about 23% maximum end deflection at the end of cooling is visible in Fig. 9, though some tests had much smaller residual deflections (see Fig. 16). This noticeable non-zero deflection could be attributed to the difference between Mf of 16 °C and ambient (or room) temperature of about 20 °C so that the forward transformation was likely to be incomplete. In particular, the actual Mf of the prestrained wires could be less than that tested with zero prestrain [18]. To verify the latter, a series of tests on the same beam were repeated in an open freezer with ambient temperature of being close to 18 °C. Fig. 10 shows a comparison of two deflection–time curves, in which the residual deflection of the test with the lower ambient temperature diminished to the only 5% maximum end deflection. This could be attributed to the increased forward transformation during the cooling period. The maximum end deflection somewhat was also reduced by the variation of ambient temperatures. Nevertheless, when both curves were normalised with reference to the respective peak deflections, the early observation was verified in Fig. 11. Indeed, the beam tested in the freezer in Fig. 10 took a little longer time to reach the same deflection in
0.6 0.4
Ambient temperature
0.2
0
of -18 C 0.0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 10. Deflection–time curves from a 130 mm carbon/epoxy beam actuated at two different ambient temperatures.
duration of 7 min. The initial linear region lasted only up to about 0.35 mm in the first 50 s. The rest of the curve exhibits a nonlinear trend and levelled off with a plateau which corresponded to the maximum end deflection of 0.89 mm. Although this level of deflection seemed very small, it could indicate that the bending moment induced by the recovery force was balanced out by the flexural rigidity of the adaptive beam. It was equally possible that non-
Fig. 11. Normalised deflection–time curves of a 130 mm carbon/epoxy beam tested at two different ambient temperatures.
2040
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041
the reverse transformation and a lesser time to cool down. To establish repeatability of the above actuation performance and consistency of the fabrication process, the entire experimentation was repeated with another 130 mm carbon/epoxy beam (B) fabricated and tested in the same way. In addition, a thermocouple was also embedded in this beam. This second adaptive beam B achieved the maximum end deflection of 0.88 mm at RT with the end deflection–time curves of both beams shown in Fig. 12. Clearly, such repeatable actuation performance from these adaptive beams confirmed once again that the present fabrication process was very adequate and reliable. Fig. 13 shows the temperature–time curve of the 130 mm carbon/epoxy beam B, in which both wire and beam surface temperatures were measured by embedded and surface-mounted thermocouples, respectively. It can be seen that the wire temperature rose almost linearly to about 195 °C in 20 s and reached the maximum temperature of about 215 °C in 50 s. It remained at that level for the rest of the heating period. Clearly, the plateau temperature was nearly four times Af of 55 °C. It implies that much higher temperature and overheating was necessary to complete the reverse transformation for prestrained wires [18] and for larger recovery stress [19,20]. The cooling response was very similar to the heating one, i.e. the wire temperature was back to the initial temperature in a similar time scale. Interestingly, the temperature on the compressive beam surface arose steadily to about 22 °C after 50 s and reached 40 °C only when cooling started. As a nominal distance between the embedded wires and the closer surface
was only about 0.25 mm, this seems to indicate that these carbon/epoxy beams were very efficient thermal insulators with the limited ability to dissipate interior heat. If such heating duration was substantial and cyclic, heat damage of epoxy could indeed become an issue. When the above temperature response was plotted against the corresponding deflection as in Fig. 14, it became clear that the deflection started only when the stress-free Af was passed and that major part of deflection took place when the plateau temperature of 215 °C was reached. In the present actuation evaluation of the adaptive beams, the examination of the effectiveness of eccentric axial load-bending transformation is extremely crucial. As the load transfer from contraction of the embedded nitinol wires to the host bending was via interfacial shear, thus the most effective region of the load transfer would be at the free end of the adaptive beams since the other end was always clamped. This interfacial shear was likely to decay towards the middle of the beam. Fig. 15 shows strain responses of two back-to-back SGs 26 mm from the clamping end. Interestingly, the strains from both SGs were of compressive nature. The peak values are 230 le from the surface remote to the wires and 455 le from the surface closer to the wires. On the basis of the superposition principle in elementary beam theory, the bending strain at the same location could be only 113 le whereas the compressive axial strain was about 343 le, which was three times greater. This seems to suggest that not all the recovery strain generated by the wire contraction was transferred to the induction of bending strain. 1.2 1.1
1.2
Heating
1.1
1.0
Cooling End deflection (mm)
End deflection (mm)
1.0 0.9 0.8 0.7 0.6
Specimen A
0.5 0.4
Specimen B
0.9 0.8 0.7 0.6 0.5 0.4 0.3
0.3
Stress-free M f
0.2
0.2
Stress-free A f
0.1
0.1
0.0
0.0 0
100
200
300
400
500
600
700
800
900
0
1000
25
50
Fig. 12. Deflection–time curves from two 130 mm carbon/epoxy beams at room temperature.
250
Heating
125
150
175
200
225
250
Fig. 14. Deflection–temperature curve from a 130 mm adaptive carbon/epoxy beam B at RT.
Cooling 100
200
50
175 150 125
-50
100
200
50
-200
200
300
500
600
700
800
900
to the nitinol wires
-300
Straingauge close to
-400
100
400
Strain gauge remote
-250 -350
25 0
300
-150
Microstrain
75
0
-100
Temperature of the beam surface closer to the wires
100
Time (second)
0
Embedded wire temperature
0
Temperature ( C)
100
0
225
0
75
Wire temperature ( C)
Time (second)
400
500
600
700
800
900
1000
Time (second)
-500 -550
Fig. 13. Temperature–time curves from a 130 mm adaptive carbon/epoxy beam B at RT.
the nitinol wires
-450
Heating
Cooling
Fig. 15. Strain–time curves from a 130 mm carbon/epoxy beam.
1000
G. Zhou, P. Lloyd / Composites Science and Technology 69 (2009) 2034–2041 2.0
Heating
1.6
End deflection (mm)
tion capability derived from the single-cycle actuation exercises was not generally suited to that for multi-cycling actuation and could result in premature failure of multi-cycled smart beams.
Cooling
1.8
2041
250-mm beam
1.4
Acknowledgement
1.2
180-mm beam
1.0
The first author acknowledges the execution of some tests by Mr. A.R. Giles.
0.8
130-mm beam
0.6
References
0.4 0.2 0.0 0
100
200
300
400
500
600
700
800
900
1000
Time (second) Fig. 16. Deflection–time curves from adaptive carbon/epoxy beams of three different lengths at room temperature.
Beams of two different additional lengths were also tested and their deflection results as included in Fig. 16 show that both the initial linear regions and the maximum end deflections seemed to be proportional to the increase in wire length, as an increase in beam length increased the electric resistance and thermal capacitance of the wires as also reported in [21]. 6. Conclusions Smart adaptive composite beams embedded with prestrained nitinol wire actuators were designed and fabricated in an autoclave with a purpose-made alignment device. The developed fabrication procedure overcame a number of technical challenges such that prestrained wires were successfully embedded eccentrically in parallel in the length direction of composite beams with three different lengths. Both single-cycle and multi-cycle thermomechanical bending actuations of these adaptive beams in the cantilever set-up were characterised experimentally by applying various levels of electric current to the nitinol wires. Such characterisation showed that the present fabrication process was repeatable and reliable. While the significant end deflections of up to 41 mm were easily achieved from the smart E-glass/epoxy beams, the limited end deflections were observed from the smart carbon/epoxy beams due primarily to our inability to find a workable solution of insulating the nitinol wires. Moreover, it seemed necessary to overheat the prestrained wires to much higher temperatures beyond the complete reverse transformation in order to generate recovery stress. The longer beams with greater actuation rates took less time to reach the same level of deflection. It was found that the actua-
[1] Birman V. Review of mechanics of shape memory alloy structures. Appl Mech Rev 1997;50:629–45. [2] Wei ZG, Sandstrom R, Miyazaki S. Shape memory materials and hybrid composites for smart system. J Mater Sci 1998;33:3763–83. [3] Baz A. Shape control of composites. In: Kelly A, Zweben C, editors. Comprehensive composite materials. London: Elsevier Science; 2000. [4] Chopra I. Review of state of art of smart structures and integrated systems. AIAA J 2002;40:2145–87. [5] Turner TL, Lach CL, Cano RJ. Fabrication and characterization of SMA hybrid composites. Proc SPIE Conf 2001:4333–60. [6] Moore CL, Bruck HA. A fundamental investigation into large strain recovery of one-way shape memory alloy wires embedded in flexible polyurethanes. Smart Mater Struct 2002;11:130–9. [7] Davis BA. Investigation of the thermomechanical response of shape memory alloy hybrid composite beams. NASA/CR-2005-213929. [8] Turner TL, Buehrle RD, Cano RJ, Fleming GA. Modeling, fabrication and testing of a SMA hybrid composite jet engine Chevron concept. J Intell Mater Syst Struct 2006;17:483–97. [9] Yang SM, Roh JH, Han JH, Lee I. Experimental studies on active shape control of composite structures using SMA actuators. J Intell Mater Syst Struct 2006;17:767–77. [10] Hebda DA, Whitlock ME, Ditman JB, White SR. Manufacturing of adaptive graphite/epoxy structures with embedded Nitinol wires. J Intell Mater Syst Struct 1995;6:220–8. [11] de Blonk BJ, Lagoudas DC. Actuation of elastomeric rods with embedded twoway shape memory alloy actuators. Smart Mater Struct 1998;7:771–83. [12] Doran CJ. The effect of prestrain on the actuation performance of embedded shape memory alloy wires. In: Proc of 2nd Euro conf on smart struct mater, Glasgow; 1994. [13] Friend CM, Morgan NB, Wise VD. Some durability issues associated with the use of shape-memory alloy ‘smart’ composites. Proc Mater Res Soc Symp 1995:513–8. [14] Zhou G, Sim LM, Brewster P, Giles AR. Through-the-thickness mechanical properties of quasi-isotropic carbon/epoxy laminates. Composites Part A 2004;35:797–815. [15] Huang W. On the selection of shape memory alloys for actuators. Mater Design 2002;23:11–9. [16] Epps J, Chandra R. Shape memory alloy actuation for active tuning of composite beams. Smart Mater Struct 1997;6:251–64. [17] Brinson LC, Huang MS, Boller C, Brand W. Analysis of controlled beam deflections using SMA wires. J Intell Mater Syst Struct 1997;8:12–25. [18] Stalman R, Tsoi K, Schrooten J. The transformation behaviour of shape memory wires embedded in a composite matrix. Proc SPIE Conf 2000;4073:88–96. [19] Brinson LC, Bekker A, Hwang S. Temperature induced deformation in shape memory alloys. Proc SPIE Conf 1994;2427:234–44. [20] Elspass WJ, Kunzmann J. Design, manufacturing and verification of laminates with embedded SMA actuators. Proc SPIE Conf 1996;2717:320–9. [21] Baz A, Inman K, McRoy J. The dynamic and thermal characteristics of shape memory actuators. Proc SPIE Conf 1994;2190:436–53.