Vibration Energy Harvesting Using Relaxor ... - Science Direct

Available online at www.sciencedirect.com

ScienceDirect Procedia Engineering 188 (2017) 432 – 439

6th Asia Pacific Workshop on Structural Health Monitoring, 6th APWSHM

Vibration energy harvesting using relaxor ferroelectric transduction Phillip Danga, Scott D. Mossa,*, Junhai Xaib, Julie M. Cairneyb b

a Aerospace Division, Defence Science and Technology Group, VIC 3207, Australia Australian Centre for Microscopy and Microanalysis, The University of Sydney, NSW 2006, Australia

Abstract The relaxor ferroelectric PIN-PMN-PT is an emerging material with outstanding electromechanical properties that holds promise for vibration energy harvesting applications. A prototype high frequency harvester was designed and assembled based on a 24PIN-PMN-PT [011] cut single crystal transducer which was mechanically loaded in the [100] direction during operation. Measured peak power output from the harvester was 78 mW from a host acceleration of 1,679 Hz and 3.5 g RMS which is a high frequency for a resonance based device. Under a wideband host acceleration of 0.5 g RMS, the harvester’s measured half power bandwidth was 619 Hz. Crown Copyright © 2016 Published by Elsevier Ltd.by This is an open © 2016 Commonwealth of Australia. Published Elsevier Ltd.access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organising committee of the 6th APWSHM. Peer-review under responsibility of the organizing committee of the 6th APWSHM Keywords: vibration energy harvesting; ferroelectric; piezoelectric; structural health monitoring

1. Introduction Structural Health Monitoring (SHM) is a tool that is becoming increasingly significant for aircraft maintenance and diagnosis, finding use with both military and civilian fleet operators [1]. Offering an alternative approach to conventional non-destructive inspection (NDI), SHM allows for the autonomous monitoring of the health of an aircraft [2]. An SHM system’s network of integrated sensors and actuators can be used to perform diagnostic

* Corresponding author. Tel.: +61-3-9626-7958; fax: +61-3-9626-7089. E-mail address: [email protected]

1877-7058 Crown Copyright © 2016 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of the 6th APWSHM

doi:10.1016/j.proeng.2017.04.505

Phillip Dang et al. / Procedia Engineering 188 (2017) 432 – 439

inspections, such as monitoring ‘hot spots’ where cracking or corrosion at a particular location on an airframe is known to be problematic [3]. The basic requirements of an SHM system are that it must be reliable, have a high probability of detection, and must satisfy airworthiness standards [2]. The system must also be rugged enough to withstand the aircraft’s operational environment, have a minimum impact on the aircraft’s systems, and ideally, should be autonomous [2]. One of the identified issues is determining the optimal means of supplying power to these in-situ SHM systems [4]. The traditional approach is to use copper wires to distribute power however that is unsatisfactory due to issues such as: (i) the additional mass of the wire, (ii) the wires occupying more space than the actual sensors, and (iii) the complicated and time consuming installation [4]. Other options such as batteries require periodic replacement or recharging [4] hence creating their own maintenance requirement [5]. An alternative powering approach is energy harvesting [6, 7]. Energy harvesting is the process of gathering energy from the surrounding environment. There are significant accelerations and strains experienced by aircraft in flight and the resultant vibration energy can be a viable source for energy harvesting devices [4]. Relaxor ferroelectric materials are a recently discovered form of piezoelectric material that can outperform traditional sintered piezoceramics by a factor of one hundred [8]. A relaxor ferroelectric that has shown potential for use in vibration energy harvesting is the single crystal material PIN-PMN-PT [8]. This material has a higher rhombohedral-tetragonal phase transition temperature of ~120qC, and twice the coercive field ~5 kV/cm, compared with earlier PMN-PT single crystal formulations [9]. Relaxor ferroelectric single crystals, also known as ternary piezoelectrics, are evolving as extremely promising materials for the design of next generation sensors, transducers and actuators owing to their outstanding electromechanical properties [9]. 2. Principle of Operation The prototype energy harvester shown in Figure 1a and b has a similar arrangement to that modelled by Du Toit [11]. The harvester design included a tungsten carbide proof mass, which was mounted on a vertically [011] cut PIN-PMN-PT crystal transducer, with mechanical to electrical energy transduction in the harvester performed by the crystal. As shown in Figure 1c the [011] direction of the crystal is considered to be the z (thickness) axis, with [Ͳͳതͳ] and [100] the x (width) and y (length) axes respectively [9]. The harvester was mounted on a 3 kg steel block (‘host mass’) and a sinusoidal host acceleration in the z direction from an electrodynamic shaker produced a mechanical load on the crystal in the [100] direction. The harvester design incorporates two neodymium disc magnets that are parallel poled and apply small compressive bias stress to the crystal. Figure 1c displays the orientation and polarisation direction of a [011] cut crystal in its rhombohedral crystal form. If the frequency of the host vibration is within the range corresponding to resonance, as determined by the effective electromechanical spring constant of the crystal and the size of the proof mass, then the electrical output from the harvester is maximised. 3. Experimental As described in Section 2 and shown in Figure 1, the prototype harvester examined in this work was assembled around a [011] cut 24PIN-PMN-PT relaxor ferroelectric crystal with dimensions 12 mm x 4 mm x 4 mm (CTG Advanced Materials). This section describes the methods used for characterising the crystal, and the vibration experiments performed with the prototype harvester. 3.1 Crystal Characterisation X-ray Diffraction (XRD) was performed on 24PIN-PMN-PT relaxor ferroelectric crystals from the same batch as that used in the prototype harvester to confirm the [011] orientation of the batch. XRD spectra were obtained using a commercial X-ray diffractometer (Shimadzu XRD S6000). The X-ray tube, with a copper (Cu) target, was operated at 40 kV and 30 mA. The divergence and scatter slits were set to 1°, and the receiving slit was set to 0.3 mm. The θ – 2θ goniometer was operated in the continuous scan mode. The scan range was 10° – 90° and the scan speed was 2°/min. In addition to XRD, impedance sweeps were performed using an impedance analyser (Solartron SI 1260). The sweeps were carried out on the pristine crystals, and also while mounted within the harvester after having been

433

434


subjected to numerous vibration experiments. Log sweeps were performed over the range 50 Hz to 10 MHz, with a 100 mV drive signal with a 1 second dwell time. Sweeps on the crystal mounted in the harvester were performed with the harvester magnetically mounted on the steel host mass. (a)

(c)

(b)

(d)

Figure 1. (a) Exploded view of the prototype vibration energy harvester, (b) photograph of harvester mounted on host mass, (c) orientation and polarisation direction of a [011] cut crystal while in rhombohedral phase, the two possible polarisation directions are indicated as [111] and [ͳതͳͳ], and (d) crystal orientation for x-ray diffraction studies of the non-electroded (Ͳͳതͳ) surface.

3.2 Harvester Response The [011] cut PIN-PMN-PT crystal transducer was mounted vertically inside a cylindrical three-dimensionally printed acrylonitrile butadiene styrene (ABS) enclosure as indicated in Figure 1a, with crystal polarisation direction as shown in Figure 1c. The enclosure was 44 mm in diameter, and 34 mm in height. The crystal was bonded between two lapped alumina discs, each 40 mm diameter and 0.8 mm thick, using adhesive (Devcon, 2-ton Epoxy) and a small compression was applied using two NdFeB disc magnets each with 20 mm diameter, 10 mm height, and 25 g mass. The alumina discs provided electrical insulation whilst also reducing the likelihood of causing permanent damage to the crystal while under mechanical load due to their high elastic modulus. The magnets provided approximately 0.9 MPa of compressive stress. This compressive stress was helpful during harvester construction since it ensured the adhesive bondlines were thin, however the compression was well below the ~20 MPa required to induce a crystal phase transformation in the material [8].


A tungsten carbide proof mass (6% cobalt binding, Carbitools) with 20 mm diameter, 20 mm height, and 91.9 g mass was mounted to the upper magnet of the harvester assembly using bees wax, increasing the total height to 54 mm. Using the lower magnet, the harvester was magnetically mounted on the steel host mass which measured 162.5 mm x 99 mm x 24.5 mm with mass ~3 kg as mentioned earlier. The host mass was attached to an electrodynamic vibration shaker (TIRAvib S514) and secured to the shaker armature using a screw, with additional adhesion provided by double sided tape which also dampened higher order host mass vibrational modes. This damping was important due to the small host mass displacements present during vibration experiments which were of the order of a few hundred nano-meters (as calculated from the measured accelerations). The centre of the block was measured and marked, ensuring that the harvester was mounted directly above the shaker armature. As per Figure 1b and Figure 2, accelerometers were used to measure accelerations of the host mass and harvester proof mass. Accelerometer outputs were measured using a digital oscilloscope (Picoscope 6407), however the host accelerometer was also connected to the vibration controller (Brüel & Kjær 7541), forming a closed loop control system. A wire was bonded (Circuitworks CW2400) to each electrode of the crystal and connected to a probe, allowing a measurement of the generated voltage. Figure 2 shows the experimental arrangement. Two types of vibration experiments were performed. The first set of experiments involved the host mass being driven by a wideband random excitation with frequencies in the range of 30 Hz to 3 kHz, with a root-mean-square (RMS) acceleration of 0.5 g. The harvester’s response to the wideband excitation was recorded over a 30 second run. Successive runs were executed across a selection of load resistances ranging from 24 kΩ through to 3,634 kΩ, with acceleration and harvester voltage outputs recorded using an oscilloscope sampling at 665k samples per second. Short circuit and open circuit conditions were also recorded to provide a basis for comparison. For the second set of experiments, the frequency was kept constant and the acceleration was ramped. A signal generator was connected directly to the power amplifier, bypassing the vibration controller, performing linear acceleration ramps from 0.3 to 3.5 g RMS at the harvester’s resonant frequency.

Figure 2. Experimental arrangement for measuring mechanical acceleration and voltage output of the harvester across load resistance R.

4. Results and Discussion The following section presents results and discussion on the characterisation of the PIN-PMN-PT crystals using X-ray diffraction and impedance analysis. Also presented are the measured harvester responses to wideband random vibration, and also single frequency acceleration ramps. Output voltage and power from the harvester are examined, as is the harvester’s wide frequency response. 4.1. Crystal Characterisation Figure 3 shows characterisation results for the PIN-PMN-PT crystals. Figure 3a shows the results of X-ray diffraction on a non-electroded side of the crystal with dimensions 4 mm x 12 mm. The (Ͳͳതͳ) surface produced

435

436


{011} and {022} reflections located at 31.35° and 65.38° respectively, indicating that the crystal structure was oriented as expected. The beam width of X-rays working on the sample ranges from millimetres to several centimetres depending on the scan angle. As shown in Figure 1 the crystals used in this work have side dimensions of 4 mm x 12 mm and it is expected that the X-ray beam covered the whole side area. Characteristic Cu Kα rays (~8 keV [10]) were used for the diffraction studies, and only these X-rays were collected by the detector. A monochromator positioned ahead of the detector removed X-rays with other wavelengths including Cu Kβ. The estimated penetration depth of Cu Kα X-rays for the PIN-PMN-PT crystal is around tens of μm. (a)

(b)

Figure 3. (a) Measured X-ray diffraction pattern from the 12 mm x 4 mm non-electroded side of the PIN-PMN-PT crystal, and (b) measured impedance sweeps of the PIN-PMN-PT crystal when new, and while mounted within the harvester and after a number of vibration studies.

Figure 3b shows the impedance sweeps of new and mounted PIN-PMN-PT crystals (post-experimentation). The pristine crystal shows a resonant peak at 47.5 kHz corresponding to the lateral fundamental resonance mode in the [100] direction. After mounting in the harvester this mode has been effectively extinguished and replaced with a peak at 1,407 Hz, which is close to the harvester’s resonant short circuit response as discussed in the next section. The reduction of the 47.5 kHz lateral mode is likely due to the clamping of the crystal after bonding within the harvester, with acoustic energy in the mode damped by adhesive at the ends of the crystal. A resonant frequency of ~1407 Hz is relatively high for a vibration energy harvester based on resonant motion, and relatively few papers have been published for harvesting from high frequency sources [6]. Higher frequency peaks near 199 kHz and 499 kHz, related to through thickness ([110] direction) or higher order resonance modes, are still present after mounting in the harvester although are somewhat reduced by the crystal clamping. 4.2. Harvester Response As described earlier, a series of 30 second runs were performed with the harvester. During each run a random waveform (30 Hz - 3 kHz) was applied to the harvester while several parameters were measured and recorded. Recorded data included the accelerations of the host and proof masses, and also the output voltage from the harvester. Load resistance was varied between runs. For each 30 second run, the response of the harvester to the random host acceleration was analysed using a 600 μs window. For each 600 μs window the RMS voltage was calculated and then converted to a power (using P = VRMS2/RL) and then averaged across the full 30 second run, providing an average power reading for the given resistive load. This procedure was repeated for all load resistances tested, generating the plot found in Figure 4 indicating the optimum load resistance was 186 kΩ.


Figure 4. Measured average power versus load resistance for wideband host vibration, 30 - 3 kHZ at an RMS acceleration level of 0.5 g.

As previously mentioned the signal from the host accelerometer was fed back into the vibration controller, which also assisted in the identifying the resonant frequencies of the harvester. Dividing the Fast Fourier Transform (FFT) of the measured proof mass acceleration by the FFT of the measured host acceleration produced the acceleration transfer functions seen in Figure 5a. Results from a selection of resistors near the 186 kΩ optimum load are displayed together with results from short circuit and open circuit conditions. Each peak in Figure 5a corresponds to the resonant frequency measured for a particular load resistor. The change in resonant frequency with resistive load is similar to that predicted by the mathematical model proposed by Du Toit et al. [11]. Figure 5a shows that the amplitude of the resonant peaks decrease as the load resistance is increased until reaching a turning point near the optimum load, after which the peaks begin increasing. The decrease in the amplitude of the mechanical resonance at optimum load is due to increased electrical output from the harvester adding electromechanical damping. In addition, as the load resistance changes from short circuit to open circuit there is a considerable shift in resonant frequency. This effect, predicted by the Du Toit model [11], is clearly shown in Figure 5b, and produces a wide frequency response near the optimum load. The stepwise appearance of Figure 5b appears as the load resistance moves from short to open circuit. A detailed comparison between the Du Toit model and the measured results will be reported elsewhere. (a)

(b)

Figure 5. (a) Measured acceleration transfer functions, using wideband excitation at an RMS acceleration of 0.5 g,with varying electrical load resistances, and (b) measured resonant frequencies identified from acceleration transfer functions versus load resistance.

437

438


Figure 6. Full width half power bandwidth as a function of load resistance, measured using a 0.5 g RMS wideband host acceleration.

(a)

(b)

Figure 7. RMS acceleration ramp from 0.3 to 3.5 g with load resistances of 186 kΩ and 300 kΩ. (a) Comparison of measured peak output voltage, and (b) comparison of measured peak output power.

Since the harvested power is proportional to the proof mass acceleration [ 12 ], the full width half power bandwidth can be estimated from the transfer function peak widths in Figure 5a. The bandwidth is plotted as a function of load resistance in Figure 6, which shows that the maximum bandwidth of 619 Hz occurs at 300 kΩ, just above optimum load resistance of 186 kΩ. As might be expected, the load resistances that produced the flatter peaks in Figure 5 also provided the widest half power bandwidths. Load resistances near short and open circuit conditions produced much narrower bandwidths as a consequence of their taller and narrower transfer functions. The harvester’s response to wideband 0.5 g RMS host vibration, shown in Figure 4, identified the optimum load resistance required however did not fully demonstrate the energy harvesting capability of the relaxor ferroelectric based harvester. To examine this, constant frequency excitation host accelerations up to an RMS acceleration of 3.5 g were used. Host acceleration was ramped from 0.3 to 3.5 g RMS in small steps. Load resistance selected for the accelerations ramps were the two best performing resistors: (i) Figure 4 indicates optimum load resistance with respect to output power was 186 kΩ, and (ii) Figure 6 shows the greatest bandwidth was achieved with a load resistance of 300 kΩ. The resonant frequencies chosen were 1,679 Hz for 186 kΩ and 1,783 Hz for 300 kΩ, both of which were identified from the transfer functions. At these frequencies and at 3.5 g RMS, the host displacement is ~400 nm, and is imperceptible to human touch. Figure 7 shows resulting voltage and power produced by the


harvester. Output voltage increased linearly as host acceleration was increased. The linearity of the response shown in Figure 7a suggests that crystal phase transformations (which potentially can lead to higher output voltages) were not occurring, and hence the electromechanical stresses being induced within the crystal during harvesting were less than the required transition stress ~20 MPa [8]. To achieve these levels of electromechanical stress the harvester would need a higher host acceleration (not possible with the apparatus used), or a high voltage bias or elevated temperature would need to be applied [8]. As shown in Figure 7a, an RMS host acceleration of 3.5 g with 186 kΩ load resistance produced a peak output voltage of 120 V whilst a 300 kΩ load resistance produced 147 V, corresponding to 78 mW and 72 mW respectively. The load resistance of 186 kΩ produced the maximum power, similar to that found earlier in Figure 4 for wideband host vibrations. 5. Conclusion To date, there has been little exploration of high frequency vibration energy harvesting using resonance based devices within the scientific literature. This paper has presented a high frequency vibration energy harvester based on a single crystal PIN-PMN-PT relaxor ferroelectric transducer. Impedance measurements suggested that the harvester’s fundamental resonant frequency was ~1,407 Hz under near short circuit conditions, high for a resonant energy harvesting device. For an RMS host acceleration of 3.5 g the optimum resistive load was found to be 186 kΩ producing 78 mW at an excitation frequency of 1,679 Hz. A 300 kΩ load resistance provided the widest bandwidth of 619 Hz. The harvester exhibited a linear response as the host acceleration was ramped to an RMS acceleration of 3.5 g, which aligns with theoretical predictions, with no evidence of electromechanical stress induced phase transformation of the crystal transducer detected. References R. Pinheiro Rulli, F. Dotta, P. A. da Silva, Flight Tests Performed by EMBRAER with SHM Systems, Key Engineering Materials 558 (2013) 305-313. [2] A. Baker, N. Rajic, C. Davis, Towards a practical structural health monitoring technology for patched cracks in aircraft structure, Composites Part A: Applied Science and Manufacturing 40 (2009) 1340-1352. [3] N. Rajic, S. Galea, Thermoelastic Stress Analysis and Structural Health Monitoring: An Emerging Nexus. Structural Health Monitoring 14 (2015) 57-72. [4] S. Moss, A. Barry, I. Powlesland, S. Galea, G. Carman, A broadband vibro-impacting power harvester with symmetrical piezoelectric bimorph-stops, Smart Materials and Structures 20 (2011) 045013. [5] G. Hart, S. Moss, D. Nagle, G. Jung, A. Wilson, C. Ung, W.K. Chiu, G. Crew, Vibration Energy Harvesting for Aircraft, Trains and Boats, Proceedings of Acoustics (2013) [6] V. Ostasevicius, V. Markevicius, V. Jurenas, M. Zilys, M. Cepenas, L. Kizauskiene, V. Gyliene, Cutting tool vibration energy harvesting for wireless sensors applications. Sensors and Actuators A: Physical 233 (2015) 310-318. [7] D. Briand, E. Yeatman, S. Roundy, Micro Energy Harvesting, John Wiley & Sons, Germany, 2015. [8] W. Dong, P. Finkel, A. Amin, K. Cunnefare, C. Lynch, Energy harvesting using the FER–FEO phase transformation in [011] cut single crystal PIN-PMN-PT, Journal of Intelligent Material Systems and Structures (2014) [9] E. Sun, S. Zhang, J. Luo, T.R. Shrout, W. Cao, Elastic, dielectric, and piezoelectric constants of Pb(In 1/2Nb1/2)O3–Pb(Mg1/3Nb2/3)O3–PbTiO3 single crystal poled along [011]c, Appl. Phys. Lett. 97 (2010) [10] X-ray Transition Energies Database, National Institute of Standards and Technology, U.S. Department of Commerce, http://physics.nist.gov/PhysRefData/XrayTrans/Html/search.html [11] N.E. Du Toit, B.L. Wardle, S-G Kim, Design Considerations for MEMS-Scale Piezoelectric Mechanical Vibration Energy Harvesters, Integrated Ferroelectrics, 71 (2005) 121-160 [12] S.A. Holmes, T.C. Green, Energy harvesting from human and machine motion for wireless electronic devices, Proc. IEEE 96 (2008) 1457–86. [1]

439