3
JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 20, NO. 6, DECEMBER 2011 1225 JMEMS Letters Nonlinear Springs for Bandwidth-Tolerant Vibration Energy Harvesting Son D. Nguyen and Einar Halvorsen Abstract—We experimentally investigate the usefulness of softening springs in a microelectromechanical systems electrostatic energy harvester under colored noise vibrations. It is shown that the nonlinear harvester has performance benefits when the vibration’s center frequency varies in the frequency range of its softening response. With a vibration 3-dB bandwidth of 50 Hz, less than 3-dB variation in output power can be obtained over a 85-Hz wide range of vibration center frequencies. Com- pared to a simulated linear-spring device, the nonlinear device gives more output power for a wide range of vibration bandwidths. The nonlinear device shows less than 1-dB variation in output power when the vibration bandwidth varies from 12 to 120 Hz and is centered on the resonant frequency. [2011-0198] Index Terms—Energy harvesting, microelectromechanical devices, non- linear systems, power generation, vibrations. I. I NTRODUCTION Traditional vibration energy harvesters are designed as linear res- onant structures. They have a very narrow bandwidth and operate efficiently only when the excitation frequency is very close to the resonant frequency of the harvesters. However, ambient vibrations may have a wide spectrum of frequencies or varying vibration spectra. There have been several attempts to overcome this limitation by tuning the resonant frequency or widening the bandwidth of the harvesters [1]. Methods have been developed to tune the resonant frequency so that it can match the excitation frequency [2], [3]. Bandwidth widening can be achieved using a generator array, a mechanical stopper [4], or bistable structures [5], [6]. Several authors have exploited nonlinear suspensions, often created with magnets, to extend the bandwidth of the harvesters by hardening [7]–[10] and/or softening nonlinearities [11], [12]. Nonlinear harvesters are usually characterized by sinusoidal excitations [9]–[13] or white noise excitations [9], [11], but real-world vibrations need not be close to these ideal cases. Vibration in a machine room is one example [14]. Of the very few works that consider finite bandwidth random vibrations [15]–[17], i.e., colored noise, only one is experimental [15]. Hence, very little is known about the potential benefits of nonlinear energy harvesting outside the idealized cases of infinite or zero bandwidth. Furthermore, the response with finite bandwidth vibrations cannot be inferred from the idealized cases because the superposition principle is not valid for a nonlinear device. In this letter, we present experiments on a nonlinear-spring harvester under colored noise vibrations over a range of vibration bandwidths and center frequencies. The device utilizes nonlinear beams in micro- machined silicon and does not rely on magnets. Except for a slightly Manuscript received June 29, 2011; revised August 27, 2011; accepted September 17, 2011. Date of publication October 27, 2011; date of current version December 2, 2011. This work was supported in part by The Research Council of Norway under Grant 191282. Subject Editor A. Seshia. The authors are with the Department of Micro and Nano Systems Technology, Faculty of Technology and Maritime Sciences, Vestfold University College, 3103 Tønsberg, Norway (e-mail: [email protected]; [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JMEMS.2011.2170824 Fig. 1. (i) Schematic drawing of MEMS electrostatic energy harvester. (ii) Spring force versus displacement of the nonlinear spring calculated by FEM and compared to a linear spring. (iii) Photograph of a corner of the electrostatic energy harvester with a nonlinear spring and part of transducer 2. smaller proof mass, it is similar to the device characterized with white noise vibrations in our previous work [11]. II. EXPERIMENTAL RESULTS Fig. 1 shows the device’s geometry and its spring force versus displacement as calculated by the finite-element method (FEM) [11]. We note that the softening spring effect occurs only in one direction, while the other direction is still hardening. The overall effect is softening, which can be seen clearly in the frequency sweeps in Fig. 2, where the broadening of the frequency response only occurs in down sweeps. This is due to the proof mass mean position shifting toward the range of lower stiffness as the amplitude increases. At a very slow sweep rate and small acceleration (inset figure in Fig. 2), the resonant frequency f 0 is about 668 Hz, and the harvester’s 3-dB bandwidth B is about 1.37 Hz. In all our experiments, the load resistors are fixed at 12.5 M(transducer 1) and 17.5 M(transducer 2), which are the optimal loads for each transducer in the linear-spring regime. This choice ensures that the loading does not artificially favor the nonlinear operation. In order to characterize the harvester under colored noise vibrations, we generate random noise vibrations with a two-sided power spectral density (PSD) given by S a (f )= 1 π A 2 f 2 (f 2 c f 2 ) 2 +∆ 2 f 2 (1) where A is the rms acceleration, f c is the center frequency of the colored noise acceleration, and is the full bandwidth (at 3 dB). A is kept fixed when comparing the average output power under different bandwidths and center frequencies of the vibrations. Fig. 3 shows the average output power versus the vibration’s center frequency f c for a vibration bandwidth of 50 Hz and at different rms accelerations. Due to the softening effect, the maximum output power is found at lower frequencies than f 0 for higher accelerations. The 3-dB bandwidth c of the responses in Fig. 3 is a measure of the tolerance toward center frequency variations of the vibrations. It is worth noting that c increases with increasing rms acceleration A 1057-7157/$26.00 © 2011 IEEE

Nonlinear Springs for Bandwidth-Tolerant Vibration Energy Harvesting

  • Upload
    einar

  • View
    213

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Nonlinear Springs for Bandwidth-Tolerant Vibration Energy Harvesting

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 20, NO. 6, DECEMBER 2011 1225

JMEMS LettersNonlinear Springs for Bandwidth-Tolerant

Vibration Energy Harvesting

Son D. Nguyen and Einar Halvorsen

Abstract—We experimentally investigate the usefulness of softeningsprings in a microelectromechanical systems electrostatic energy harvesterunder colored noise vibrations. It is shown that the nonlinear harvesterhas performance benefits when the vibration’s center frequency variesin the frequency range of its softening response. With a vibration 3-dBbandwidth of 50 Hz, less than 3-dB variation in output power can beobtained over a 85-Hz wide range of vibration center frequencies. Com-pared to a simulated linear-spring device, the nonlinear device gives moreoutput power for a wide range of vibration bandwidths. The nonlineardevice shows less than 1-dB variation in output power when the vibrationbandwidth varies from 12 to 120 Hz and is centered on the resonantfrequency. [2011-0198]

Index Terms—Energy harvesting, microelectromechanical devices, non-linear systems, power generation, vibrations.

I. INTRODUCTION

Traditional vibration energy harvesters are designed as linear res-onant structures. They have a very narrow bandwidth and operateefficiently only when the excitation frequency is very close to theresonant frequency of the harvesters. However, ambient vibrationsmay have a wide spectrum of frequencies or varying vibration spectra.There have been several attempts to overcome this limitation by tuningthe resonant frequency or widening the bandwidth of the harvesters[1]. Methods have been developed to tune the resonant frequency sothat it can match the excitation frequency [2], [3]. Bandwidth wideningcan be achieved using a generator array, a mechanical stopper [4], orbistable structures [5], [6]. Several authors have exploited nonlinearsuspensions, often created with magnets, to extend the bandwidth ofthe harvesters by hardening [7]–[10] and/or softening nonlinearities[11], [12]. Nonlinear harvesters are usually characterized by sinusoidalexcitations [9]–[13] or white noise excitations [9], [11], but real-worldvibrations need not be close to these ideal cases. Vibration in a machineroom is one example [14]. Of the very few works that consider finitebandwidth random vibrations [15]–[17], i.e., colored noise, only oneis experimental [15]. Hence, very little is known about the potentialbenefits of nonlinear energy harvesting outside the idealized casesof infinite or zero bandwidth. Furthermore, the response with finitebandwidth vibrations cannot be inferred from the idealized casesbecause the superposition principle is not valid for a nonlinear device.

In this letter, we present experiments on a nonlinear-spring harvesterunder colored noise vibrations over a range of vibration bandwidthsand center frequencies. The device utilizes nonlinear beams in micro-machined silicon and does not rely on magnets. Except for a slightly

Manuscript received June 29, 2011; revised August 27, 2011; acceptedSeptember 17, 2011. Date of publication October 27, 2011; date of currentversion December 2, 2011. This work was supported in part by The ResearchCouncil of Norway under Grant 191282. Subject Editor A. Seshia.

The authors are with the Department of Micro and Nano SystemsTechnology, Faculty of Technology and Maritime Sciences, VestfoldUniversity College, 3103 Tønsberg, Norway (e-mail: [email protected];[email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JMEMS.2011.2170824

Fig. 1. (i) Schematic drawing of MEMS electrostatic energy harvester.(ii) Spring force versus displacement of the nonlinear spring calculated by FEMand compared to a linear spring. (iii) Photograph of a corner of the electrostaticenergy harvester with a nonlinear spring and part of transducer 2.

smaller proof mass, it is similar to the device characterized with whitenoise vibrations in our previous work [11].

II. EXPERIMENTAL RESULTS

Fig. 1 shows the device’s geometry and its spring force versusdisplacement as calculated by the finite-element method (FEM) [11].We note that the softening spring effect occurs only in one direction,while the other direction is still hardening. The overall effect issoftening, which can be seen clearly in the frequency sweeps in Fig. 2,where the broadening of the frequency response only occurs in downsweeps. This is due to the proof mass mean position shifting towardthe range of lower stiffness as the amplitude increases. At a very slowsweep rate and small acceleration (inset figure in Fig. 2), the resonantfrequency f0 is about 668 Hz, and the harvester’s 3-dB bandwidth Bis about 1.37 Hz. In all our experiments, the load resistors are fixedat 12.5 MΩ (transducer 1) and 17.5 MΩ (transducer 2), which arethe optimal loads for each transducer in the linear-spring regime. Thischoice ensures that the loading does not artificially favor the nonlinearoperation.

In order to characterize the harvester under colored noise vibrations,we generate random noise vibrations with a two-sided power spectraldensity (PSD) given by

Sa(f) =1

π

∆A2f2

(f2c − f2)2 + ∆2f2

(1)

where A is the rms acceleration, fc is the center frequency of thecolored noise acceleration, and ∆ is the full bandwidth (at 3 dB). A iskept fixed when comparing the average output power under differentbandwidths and center frequencies of the vibrations.

Fig. 3 shows the average output power versus the vibration’s centerfrequency fc for a vibration bandwidth of 50 Hz and at different rmsaccelerations. Due to the softening effect, the maximum output poweris found at lower frequencies than f0 for higher accelerations. The3-dB bandwidth ∆c of the responses in Fig. 3 is a measure of thetolerance toward center frequency variations of the vibrations. It isworth noting that ∆c increases with increasing rms acceleration A

1057-7157/$26.00 © 2011 IEEE

Page 2: Nonlinear Springs for Bandwidth-Tolerant Vibration Energy Harvesting

1226 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 20, NO. 6, DECEMBER 2011

Fig. 2. Linearly increasing and decreasing frequency sweeps for rms acceler-ations of 0.045, 0.064, 0.088, 0.119, and 0.150 g. (Dot-and-dash lines) Downsweeps. (Solid lines) Up sweeps. The bias is 36 V.

Fig. 3. Average output power versus vibration center frequency for ∆ =50 Hz and rms accelerations of 0.097, 0.231, 0.360, and 0.488 g. Eachaverage output power point is obtained by averaging the rms values of threeexperiments. Each of these experiments lasts for 40 s.

beyond a certain value. For A = 0.097 g, ∆c ≈ 50 Hz, which is equalto the vibration’s bandwidth ∆. For larger A, ∆c > ∆, i.e., ∆c = 60,75, and 85 Hz for A = 0.231, 0.360, and 0.488 g, respectively. If theresponse bandwidth is taken instead at 1 dB below the peak power, itsvalues are 30, 31, 40, and 50 Hz, respectively, for the four values of A.The 1-dB bandwidth of the vibration is 25.4 Hz.

The output PSD as a function of frequency for some selectedvibration center frequencies fc at A = 0.488 g is shown in Fig. 4. Forfc = 660 Hz, which is close to f0, the harvester bandwidth is about48 Hz. When fc moves to lower frequencies, the harvester’s bandwidthis wider (about 63 Hz for fc = 640 and 72 Hz for fc = 620 Hz) butstill keeps the high cutoff frequency around the resonant frequency. Atfc = 600 Hz, the harvester’s bandwidth reduces because of the low-ered spectral weight of the vibration in the sensitive frequency rangeof the harvester. The harvester responds with a narrow bandwidtharound the resonant frequency when the frequency is further reducedto fc = 570 Hz because the frequency mismatch is too large to excitethe harvester into the strongly nonlinear regime of softening springbehavior. Although the peak value of the output PSD may appear high,the integrated PSD is not. It is proportional to the output power inFig. 3, where we obtained 26.9 nW for fc = 570 Hz, which is 7.2 dBlower than 140.3 nW obtained for fc = 660 Hz.

Next, we study the effect of the vibration bandwidth ∆ on the av-erage output power. The average output power against ∆ for different

Fig. 4. Output PSD versus frequency for different vibration center frequen-cies, vibration bandwidth ∆ = 50 Hz, and rms acceleration A = 0.488 g.

Fig. 5. Average output power versus vibration bandwidth ∆ for differentvibration center frequencies fc and an rms acceleration of 0.010 g.

fc’s and very small rms accelerations is shown in Fig. 5. At this level ofacceleration, the contribution of the spring nonlinearity is insignificant,and hence, the harvester operates in a linear-spring regime. It canbe seen that the harvester then works best when fc is close to f0 =668 Hz and ∆ < B [18]. The output power significantly reduces whenfc is far from f0 (i.e., fc = 630 and 610 Hz) or when ∆ > B because,then, much of the spectral content of the vibration is filtered out. Thatis the main limitation of energy harvesters with linear springs.

Fig. 6 shows the variations of the average output power with thevibration’s bandwidth ∆ for an rms acceleration of 0.488 g, whichis large enough to give the spring softening effect. We also plot thesimulated average output power for a harvester using linear springsunder the same excitation. When fc = f0 = 668 Hz and ∆ is small,which is the condition for the best performance of a linear harvester,the average output power with nonlinear springs is much lower thanthat with a linear spring. The linear spring would result in 1.5 µWat ∆ = 0.2 Hz with the proof mass reaching the maximum possibledisplacement, i.e., hits the mechanical end stops (not shown in Fig. 6).However, when ∆ is larger than about 80 Hz, the average power ofthe linear spring harvester is smaller than the average power of thenonlinear-spring harvester. When fc shifts to lower frequencies, i.e.,at 630 or 610 Hz, the performance of the harvester using nonlinearsprings is much better than the harvester using linear springs, alwaysgiving higher average power than the linear-spring harvester. In partic-ular, at fc = 630 Hz and ∆ = 30 Hz, the average output power withnonlinear springs is higher than that with linear springs by a factor of3.8. The output power for the nonlinear device varies less than 1 dBover a range of bandwidths from 12 to 120 Hz for fc = 668 Hz.

Page 3: Nonlinear Springs for Bandwidth-Tolerant Vibration Energy Harvesting

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 20, NO. 6, DECEMBER 2011 1227

Fig. 6. Measured average output power of nonlinear-spring device versusvibration bandwidth ∆ for different center frequencies, fc = f0 = 668 Hz,fc = 630 Hz and fc = 610 Hz, and an rms acceleration of 0.488 g. SPICEsimulations for a comparable linear-spring harvester.

III. CONCLUSION

We have experimentally investigated the usefulness of softeningsprings in a microelectromechanical systems (MEMS) vibration en-ergy harvester under colored noise excitations. Varying the vibrationcenter frequency, we found a response bandwidth of up to 85 Hz,exceeding the vibration bandwidth of 50 Hz. For vibration centerfrequencies 38 or 58 Hz below the harvester’s resonant frequency,the nonlinear-spring harvester always achieves more power than alinear-spring harvester regardless of the vibration’s bandwidth. Thenonlinear device showed less than 1-dB variation in output powerwhen the vibration bandwidth varied from 12 to 120 Hz around theresonant frequency. In conclusion, the nonlinear device is toleranttoward variations in vibration center frequency and bandwidth. It ispromising that the desired nonlinearity can be obtained purely bymechanical design.

REFERENCES

[1] D. Zhu, J. Tudor, and S. Beeby, “Strategies for increasing the operatingfrequency range of vibration energy harvesters: A review,” Meas. Sci.Technol., vol. 21, no. 2, p. 022 001, Feb. 2010.

[2] E. S. Leland and P. K. Wright, “Resonance tuning of piezoelectric vi-bration energy scavenging generators using compressive axial preload,”Smart Mater. Struct., vol. 15, no. 5, pp. 1413–1420, Oct. 2006.

[3] L. Gu and C. Livermore, “Passive self-tuning energy harvester for ex-tracting energy from rotational motion,” Appl. Phys. Lett., vol. 97, no. 8,pp. 081 904-1–081 904-3, Aug. 2010.

[4] M. Soliman, E. M. Abdel-Rahman, E. F. El-Saadany, and R. R. Mansour,“A design procedure for wideband micropower generators,” J. Micro-electromech. Syst., vol. 18, no. 6, pp. 1288–1299, Dec. 2009.

[5] F. Cottone, H. Vocca, and L. Gammaitoni, “Nonlinear energy harvesting,”Phys. Rev. Lett., vol. 102, no. 8, p. 080 601, Feb. 2009.

[6] A. Erturk, J. Hoffmann, and D. Inman, “A piezo-magneto-elastic structurefor broadband vibration energy harvesting,” Appl. Phys. Lett., vol. 94,no. 25, pp. 254 102-1–254 102-3, Jun. 2009.

[7] S. G. Burrow and L. R. Clare, “A resonant generator with non-linearcompliance for energy harvesting in high vibrational environments,” inProc. IEEE IEMDC, 2007, vol. 1, pp. 715–720.

[8] B. Marinkovic and H. Koser, “Smart sand—A wide bandwidth vibra-tion energy harvesting platform,” Appl. Phys. Lett., vol. 94, no. 10,pp. 103 505-1–103 505-3, Mar. 2009.

[9] L. G. W. Tvedt, D. S. Nguyen, and E. Halvorsen, “Nonlinear behavior ofan electrostatic energy harvester under wide- and narrowband excitation,”J. Microelectromech. Syst., vol. 19, no. 2, pp. 305–316, Apr. 2010.

[10] M. Marzencki, M. Defosseux, and S. Basrour, “MEMS vibration en-ergy harvesting devices with passive resonance frequency adaptationcapability,” J. Microelectromech. Syst., vol. 18, no. 6, pp. 1444–1453,Dec. 2009.

[11] D. S. Nguyen, E. Halvorsen, G. U. Jensen, and A. Vogl, “Fabricationand characterization of a wideband MEMS energy harvester utilizingnonlinear springs,” J. Micromech. Microeng., vol. 20, no. 12, p. 125 009,Dec. 2010.

[12] S. C. Stanton, C. C. McGehee, and B. P. Mann, “Reversible hysteresisfor broadband magnetopiezoelastic energy harvesting,” Appl. Phys. Lett.,vol. 95, no. 17, pp. 174 103-1–174 103-3, Oct. 2009.

[13] S. C. Stanton, C. C. McGehee, and B. P. Mann, “Nonlinear dynamics forbroadband energy harvesting: Investigation of a bistable piezoelectric in-ertial generator,” Phys. D, Nonlinear Phenom., vol. 239, no. 10, pp. 640–653, May 2010.

[14] L. M. Miller, E. Halvorsen, T. Dong, and P. K. Wright, “Modeling and ex-perimental verification of low-frequency MEMS energy harvesting fromambient vibrations,” J. Micromech. Microeng., vol. 21, no. 4, p. 045 029,Apr. 2011.

[15] D. A. W. Barton, S. G. Burrow, and L. R. Clare, “Energy harvesting fromvibrations with a nonlinear oscillator,” J. Vib. Acoust., vol. 132, no. 2,pp. 021 009-1–021 009-7, Apr. 2010.

[16] M. F. Daqaq, “Response of uni-modal duffing-type harvesters to ran-dom forced excitations,” J. Sound Vib., vol. 329, no. 18, pp. 3621–3631,Aug. 2010.

[17] A. M. Wickenheiser and E. Garcia, “Power optimization of vibrationenergy harvesters utilizing passive and active circuits,” J. Intell. Mater.Syst. Struct., vol. 21, no. 13, pp. 1343–1361, Sep. 2010.

[18] E. Halvorsen, “Energy harvesters driven by broadband random vi-brations,” J. Microelectromech. Syst., vol. 17, no. 5, pp. 1061–1071,Oct. 2008.