Energy harvesting of radio frequency and vibration energy to enable wireless sensor monitoring of civil infrastructure

Transcription

1 Energy harvesting of radio frequency and vibration energy to enable wireless sensor monitoring of civil infrastructure Tzeno Galchev, James McCullagh, Rebecca L. Peterson, Khalil Najafi*, Amir Mortazawi Electrical and Computer Engineering, University of Michigan, 1301 Beal Avenue, Ann Arbor, MI USA ABSTRACT To power distributed wireless sensor networks on bridges, traditional power cables or battery replacement are excessively expensive or infeasible. This project develops two power harvesting technologies. First, a novel parametric frequency-increased generator (PFIG) is developed. The fabricated PFIG harvests the non-periodic and unprecedentedly low frequency (DC to 30 Hz) and low acceleration ( m/s 2 ) mechanical energy available on bridges with an average power > 2 µw. Prototype power conversion and storage electronics were designed and the harvester system was used to charge a capacitor from arbitrary bridge-like vibrations. Second, аn RF scavenger operating at medium and shortwave frequencies has been designed and tested. Power scavenging at MHz frequencies allows for lower antenna directivities, reducing sensitivity to antenna positioning. Furthermore, ambient RF signals at these frequencies have higher power levels away from cities and residential areas compared to the UHF and SHF bands utilized for cellular communication systems. An RF power scavenger operating at 1 MHz along with power management and storage circuitry has been demonstrated. It powers a LED at a distance of 10 km from AM radio stations. Keywords: Vibration Harvesting, PFIG, Parametric Generator, Frequency Up-Conversion, Bridge Vibrations, RF Harvesting, AM Band Harvesting, RF Energy Scavenging 1. INTRODUCTION Buildings, highways, bridges, dams, and railways form the backbone of our societies. The health and performance of some of these systems are severely undermanaged. One category of civil structures that typically features ambitious engineering challenges is bridges. In the United States highway bridges undergo a visual inspection every two years [1]. This constitutes gross mismanagement in the face of other statistics. For example, as of December 2009 the US Department of Transportation rates 71,179 bridges as structurally deficient and 78,468 as functionally obsolete [2], which constitutes 25% of the 603,254 bridges in total. Of course, epic failures in these structures such as the 2007 I-35W bridge collapse in Minnesota garner a great deal of media attention. However, it is not widely known that between 1989 and 2000 there were 503 bridge collapses in the United States [3]. Closer monitoring of these structures is mainly limited by economics. Monitoring 600,000 bridges by human observation is prohibitively expensive. Efforts to develop automated monitoring systems are ongoing [4]. For example, accelerometers have been suggested to monitor changes to mode shapes or damping characteristics. Other methods are also under consideration. However, even these automated systems can be costly. One of the main challenges rests in the wires used to route power and data from the sensors to the processing point, because wires are physically vulnerable and expensive to install and to maintain [5]. In fact, wires can cause a tenfold increase in the cost of a sensor. For this reason interest in wireless sensing systems has grown, because such systems offer the promise of improved structural health monitoring and management. However, wireless sensors need wireless power to make them economically viable. This makes energy harvesting technology very important in these systems. Batteries and other stored energy means can be used however multiple replacements will be needed through the lifetime of the system. Each replacement will carry with it a significant cost mainly due to labor. Since hundreds to thousands of sensors would be needed on typical bridges, the cost of non-renewable energy sources is a formidable barrier to the adoption of wireless monitoring systems. *najafi@umich.edu; phone +1 (734) ; fax +1 (734) ; V. 1 (p.1 of 9) / Color: No / Format: Letter / Date: :31:54 PM

2 This paper presents a hybrid renewable power source for structural health monitoring wireless sensing systems, in particular for bridges. The power source (Figure 1) seeks to harvest energy from traffic-induced vibrations as well as radio frequency (RF) energy surrounding a typical bridge. Both energy sources have specific characteristics in the context of bridges, and these present a number of engineering challenges that must be overcome. The vibrations found on a typical bridge have a very low-amplitude and are largely non-periodic, with frequencies in the DC-30 Hz range. These characteristics make it challenging to harvest and convert the vibration energy into electricity, but vibrations can be found in abundance on many bridges due to passing traffic. A novel Parametric Frequency-Increased Generator (PFIG) harvester has been designed [6] in which a mass, moving with bridge vibrations, latches sequentially onto two electro-magnetic transducers which, when released, oscillate at a higher frequency, increasing the harvester bandwidth and mechanical-to-electrical conversion efficiency. The PFIG generator has been shown to have 10x better efficiency over other reported harvesters in this space [7], due to its low-frequency non-resonant operation. Additionally, an RF scavenger operating at medium and shortwave frequencies has been designed and tested. RF signals from high power broadcast stations (AM, FM, and TV) are available nationwide and they allow for lower antenna directivity, reducing the sensitivity to antenna positioning. An RF power scavenger operating at 1 MHz along with power management and storage circuitry has been demonstrated. The detailed design and test results will be presented for both types of harvesters. Figure 1. Architecture of the hybrid power source for wireless bridge health monitoring applications. (а) (b) Figure 2. Vibration data from two popular bridge types: (a) a steel girder-concrete deck composite structure in Ypsilanti, MI and (b) a suspension bridge in Valejo, California. The measured acceleration as well as the frequency response of a representative sensor location on each bridge is shown. The traffic-induced vibrations have a very small amplitude and a variable low-frequency response V. 1 (p.2 of 9) / Color: No / Format: Letter / Date: :31:54 PM

3 2. HARVESTING TRAFFIC-INDUCED BRIDGE VIBRATIONS 2.1 Bridge vibration characteristics The vibration characteristics of two different types of bridges were measured in order to determine the technical specifications for the mechanical harvester. Data from a typical highway flyover steel girder-concrete deck composite structure, Figure 2a, and from a nearly kilometer-long suspension bridge, Figure 2b, were collected and analyzed. A triaxial accelerometer (Crossbow CXL02TG3) was sampled at 100 Hz. The girder bridge is the Grove Street (GS) I-94 flyover in Ypsilanti, Michigan. There were 20 measurement points on the bridge. The suspension bridge is the New Carquinez Bridge (NC) in Valejo, California and it had 11 sensor locations. Acceleration recordings were taken for each node under routine traffic loads. The waveforms shown in Figure 2 are very typical of the remaining sensor locations. The GS bridge had peak accelerations in the range of mg (1mg = 9.8 x 10-3 m/s 2 ). The Discrete Fourier Transform of a portion of these data is also shown in Figure 2 revealing that the vibrations are fairly arbitrary. There is no identifiable peak. The spectral content is mostly spread out in the DC-30 Hz range. The remaining sensor locations confirm the arbitrary nature of the vibrations. The frequency spectrum for each sensor looks different, although most of the energy is again contained within the DC-30 Hz range. Data from the NC bridge are similar in that from sensor to sensor the frequency response is different. However, the suspension bridge exhibits larger accelerations, as high as 130 mg in some places. In order to design a mechanical harvester capable of operation on different types of bridges without redesign it should work with extremely low accelerations, mg maximum. Additionally, the data shows that a widebandwidth vibration harvester is needed due to the variable frequency characteristics between the two types of bridges and between different sensor locations on the same bridge. Figure 3. a) Parametric Frequency Increased Generator (PFIG) architecture b) Illustration of the method of operation [8]. 2.2 Mechanical harvester design Typical vibration harvesters use a suspended mass that moves in response to external forces. A transducer damps this motion and converts it into electrical energy. The maximum output power density scales proportionally with the input acceleration and frequency. Inertial generators are usually operated at their resonant frequency in order to take advantage of the inherent mechanical amplification due to the device s quality factor. Unfortunately, this technique is not effective for harvesting bridge vibrations because as shown in the last section they do not have a well defined period and both the acceleration and frequency are low for bridges. This makes energy conversion in these environments particularly challenging. To surmount these challenges, a broadband Parametric Frequency Increase Generator (PFIG) was designed [6] and an implementation suitable for harvesting bridge vibrations was presented [8]. Figure 3 illustrates the PFIG cross-section and the method of operation. The PFIG consists of a suspended mass that moves in response to the applied external displacement on the frame, triggering higher-frequency mechanical resonance of a frequency V. 1 (p.3 of 9) / Color: No / Format: Letter / Date: :31:54 PM

4 increased generator (FIG). Two FIGs are used: one above and one below the mass. The mass sequentially and repeatedly actuates the two FIGs by briefly latching to them using permanent NdFeB magnets. Using electromagnetic transduction, the FIGs convert the mechanical energy imparted on them by the inertial mass into electricity. The PFIG structure serves a dual purpose. First, due to its non-resonant operation, it decouples the harvester operating frequency from that of the ambient environment. This feature allows the PFIG to operate over a broad frequency range. Second, the electromechanical coupling (damping force) will be weaker as the frequency (and thus velocity) drops. The FIG component of the generator gets its name from a concept called frequency up-conversion [9, 10], a method to increase the effectiveness of low-frequency scavengers. This is achieved by implementing a mechanical conversion, such that the internal operating frequency of the generator is increased over the input frequency. The damping force is thereby scaled proportionately. The FIGs operate at a frequency that is an order of magnitude higher than the ambient vibration. Figure 4. An illustration of the PFIG shown alongside a photograph of the partially opened device is used to describe the layout of the components within the generator enclosure. The FIGs are show on top and bottom, clamped using setscrews inside a holding fixture. In between, the WC inertial mass is suspended on two copper springs. Figure 5. Photograph of the fabricated PFIG generator. The generator is compared with a standard D battery. 2.3 Implementation and results The implementation of the PFIG is shown in Figure 4. An illustration of the device structure is shown next to a photograph of the fabricated harvester, highlighting its internal components. The generator is housed within an aluminum enclosure. A large tungsten carbide (WC) inertial mass can be seen in the middle suspended on either side using copper springs. The two FIGs (top and bottom) surround the inertial mass. The FIG hardware is contained within a movable enclosure that is held in place by set screws. The built in motion range is necessary in order to bias the PFIG against the effects of gravity. The springs for both the FIG and the inertial mass are fabricated out of 250 μm thick copper alloy. NdFeB magnets are adhered to the FIG springs. Coils are wound from 50 μm diameter enameled copper wire around custom machined aluminum bobbins and mechanically stabilized within the transducer compartment. The finished PFIG measures 3.3 cm in diameter and is 7.3 cm tall. The internal volume of the device, featuring all of the transduction mechanisms, the inertial mass, and all of the space needed for the components to move is 43 cm 3. The finished device can be seen in Figure 5 where the PFIG is shown alongside a standard D size battery. The harvester was tested on an APS Dynamics APS113 long stroke linear shaker. Each FIG was loaded with a 1.5 kω resistor in order to match its output impedance. The fabricated harvester was able to function from accelerations as low as 55 mg, putting it on the upper end of the design target. Figure 6 shows the voltage generated by each FIG across the load resistor while the PFIG is excited by a sinusoidal excitation at 5 Hz. The operation of the PFIG can be better V. 1 (p.4 of 9) / Color: No / Format: Letter / Date: :31:54 PM

5 understood by examining the voltage waveform. The sequential excitation of the FIGs due to the inertial mass attachment and subsequent detachment from each FIG can easily be recognized. The lowest frequency measurement that the test setup allowed was 2 Hz, at which the PFIG was able to generate 2.3 μw of average power (57 μw peak power) at 55 mg acceleration. The frequency response was measured while the harvester was exposed to accelerations in the range of g (Figure 7). This constitutes an unprecedented operation range for a vibration harvester: almost two full orders of magnitude. These measurements were performed without any modifications or tuning to the PFIG. Voltage (V) Voltage (V) FIG Mass path Mass attaches FIG 2 Power 0.2 generation Time (s) Figure 6. Oscilloscope recording showing the parametric generator operation from 55 mg at 5 Hz. The top and bottom voltage waveforms correspond to the top and bottom FIG devices as the inertial mass snaps back and forth between them [8]. Average Power (uw) g 0.3g 1g Frequency (Hz) Figure 7. Frequency response of the PFIG generator at different vibration amplitudes. The PFIG generator is able to operate over an unprecedented acceleration range of nearly 2 orders of magnitude [8]. Figure 8. Harvester power conversion system. Each FIG is connected to a 6-stage Cockroft-Walton multiplier. The output of these multipliers is cascaded and stored on a capacitor. Figure 9. The harvester system is tested on a shaker table driven by a signal mimicking bridge vibrations. The figure shows the voltage on the storage capacitor as the harvester charges it. 2.4 Interface electronics Appropriate power conversion electronics have to be designed to turn the decaying alternating (AC) voltage into a usable constant (DC) voltage level. The voltage levels produced by the PFIG are relatively low and so a boosting scheme is essential in this application. A charge pump approach is used because it serves a dual purpose of both rectifying and V. 1 (p.5 of 9) / Color: No / Format: Letter / Date: :31:54 PM

6 boosting the voltage produced by each FIG. The design of the power conversion system is shown in Figure 8. A 6 stage Cockcroft-Walton (CW) multiplier is attached to each of the two FIGs. Schottky diodes (BAT54WS) and 10 μf capacitors are used to construct the multiplier stages. The outputs of the two multipliers are cascaded to further increase the voltage and to combine the two outputs into one. The resulting charge is stored on a 47 μf capacitor. 2.5 System Testing The harvester system is tested on a shaker table by subjecting it to bridge-like arbitrary vibrations in order to demonstrate operation from random (non-periodic) excitation. A 20 second sample of the bridge acceleration data is looped and replayed by the shaker. The acceleration profile can be seen in the top of Figure 9. The voltage on the storage capacitor can be seen to rise in the bottom plot of Figure 9 as the harvester converts and stores mechanical energy. Some ripples can be seen on the rising voltage. The ripples are a result of parasitic discharging in the multipliers and the storage capacitor. During the portions of time where there is a gap between large acceleration spikes the parasitic discharging dominates. This occurs because this first version of the power conversion electronics is implemented on a breadboard using off the shelf components and there is limited control over parasitics. 3. MEDIUM AND SHORT WAVE FREQUENCY RF HARVESTING Radio frequency (RF) energy scavenging circuits can be used to scavenge energy from the ambient RF broadcast spectrum as an alternative source of energy. RF signals from high power broadcast stations AM (530~1700 khz), FM (88~108 MHz), TV (54~890 MHz) can be found nationwide. This availability renders broadcast stations as a promising source of energy for unattended sensor nodes deployed to the field. In this study, the AM radio spectrum is selected given its ubiquitous availability. The key challenges for the development of RF energy harvesters are: 1) efficient yet small antennas to pick up the stray electromagnetic signals available from radio as well as mobile phone stations; 2) efficient zero threshold rectifier circuits along with DC to DC converters and matching circuitry for proper coupling to the aforementioned antennas. The first topic is covered in greater detail in this paper, while the design of the DC to DC converter was examined in [3]. Figure 10. Block diagram of the RF energy harvester. 3.1 Antenna design and impedance matching The block diagram of the RF energy harvester is shown in Figure 10. The main blocks of the energy harvester include an antenna, rectifier circuit, boost converter or charge pump, and an energy storage element. RF energy captured by the antenna is delivered to a rectifier through a matching circuit to maximize its efficiency. The RF signal is then converted to a DC voltage which charges up an energy storage capacitor. When the capacitor s stored charge exceeds a certain threshold level, it powers up a DC to DC boost converter that generates a voltage that is high enough for sensor operation. Use of the AM frequency band is an attractive approach to more commonly investigated energy harvesting from RF sources at much higher frequencies such as WiFi sources. AM radio stations provide wide coverage and the transmitter and receiver antennas do not need to be in line of sight of each other. However, due to their large electrical wavelengths, design of compact antennas with reasonable efficiencies is very challenging. Ferrite based antennas need to be employed to enhance the magnetic flux while maintaining a compact size. The radiation resistance of the ferrite core antennas can be represented by: V. 1 (p.6 of 9) / Color: No / Format: Letter / Date: :31:54 PM

7 R rad 2 2 NA = 31200μer λ (1) where N is the number of windings, μer is the effective permeability of the ferrite rod, A is the cross section area of the ferrite material and λ is the free space wavelength of the AM signal to be harvested [11-13]. At the same time, the antenna loss is determined by: Rloss = Rf + RDC + RAC = ( tanδ f + tanδdc + tan δac ) ωl (2), where Rf, RDC, RAC are resistances due to the ferrite core, DC wiring, and the AC proximity effect. Antenna losses must be minimized in order to optimize its efficiency. Various loss mechanisms can be quantified by defining loss tangent values for them. The DC loss tangent is given by: tanδ dc 9 4ρclw10 = ω A Nnπ d L 2 (3), 2 where A ( / ) L nh turn is a geometry coefficient determined by measurement, ρ c is the resistivity of the winding material, l w is the circumference of each winding, and n and d are the number of strands and diameter of each strand of the multi-stranded litz-wire used to design the antenna. The AC loss tangent is given by: ke fnnd tanδ ac = A L 4, (4) where k E is geometry dependent proximity effect coefficient determined experimentally. The ferrite loss tangent is determined based on the choice of ferrite core material. In general, given the ferrite rod geometry, the efficiency of the antenna can be maximized by optimizing its quality factor Q. A typical quality factor of approximately 900 is achieved for the ferrite antenna designed for AM band power harvesting. Various low threshold rectifiers were used to convert the RF signal to a DC voltage. Large signal analysis was performed in order to accurately determine the rectifier s dynamic impedance. The L-type matching circuit (Figure 11), consisting of the inductance of ferrite antenna connected in shunt with a variable capacitor, allows the antenna radiation resistance to be matched to the rectifier s impedance in order to optimize its efficiency. Z L ZD = R jx X Q = R Q CD = ωr D R = R + Q D 2 (1 ) Figure 11. Illustration of the impedance matching between the ferrite antenna and the diode rectifier V. 1 (p.7 of 9) / Color: No / Format: Letter / Date: :31:54 PM

8 Figure 12. Detailed Diagram of the AM power harvester. 3.2 АM band harvester results A more detailed diagram of the AM band energy harvesting circuit is shown in Figure 12. The output of the rectifier is connected to a super capacitor for energy storage. When the voltage across the super capacitor exceeds a predetermined value set by the sensing circuit, the DC to DC voltage converter is activated. The converter circuit generates voltages in excess of 2V, sufficient enough to power an LED for a small period of time. The LED is used only for demonstration purposes; in real applications the circuit would be used to power a wireless sensor. A rechargeable battery is used within the circuit and it is also charged during the operation of the DC to DC boost converter. The measured efficiency of the DC to DC converter circuit as a function of input power is shown in Figure 13. The DC to DC converter circuit can achieve efficiencies as high as 70% with sufficient input power levels. Finally a graph of the energy harvested by the AM power scavenger as a function of its distance from an AM radio station with a typical transmit power level of 10 kw at 1 MHz is shown in Figure 14. As shown in Figure 14, energy levels of approximately 100μJ can be harvested at a distance of 10km from the AM station assuming that a large enough capacitor is used to store the energy. Future work in the areas of RF power harvesting will include the development of more efficient antenna and rectifier circuits that would allow power harvesting at larger distances from the AM transmitters. Figure 13. Measured efficiency of the DC to DC converter circuit as a function of the input power level. Figure 14. Stored energy as a function of AM power scavenger s distance to an AM station. 4. CONCLUSION The structural health of large infrastructure systems can be monitored using wireless sensor networks. However, in order to make even automated wireless monitoring cost-effective, the use of energy harvesting technology is essential. This paper discussed the component design of a hybrid power source capable of converting both vibrations and RF V. 1 (p.8 of 9) / Color: No / Format: Letter / Date: :31:54 PM

9 energy for use in wireless bridge health monitoring. The development of a novel Parametric Frequency Increased Generator (PFIG) for harvesting traffic-induced bridge vibrations was discussed. The vibrations found on bridges have small amplitudes and they are low frequency and non-periodic. The developed harvester is shown to be capable of efficiently converting these vibrations to electrical energy. It can function from accelerations as low as 0.54m/s 2 over a frequency band of DC-18Hz. A power conversion system was developed and a capacitor was charged from arbitrary bridge vibrations. Additionally, a radio frequency power harvesting device was proposed that can draw power from AM radio waves. AM broadcasts can be found outside of cities and other urban centers, they have high broadcast power, and they reduce the requirement for transmitter/receiver line of sight. An RF power scavenger operating at 1MHz was developed. Appropriate power management and storage circuitry was also demonstrated. It powers a LED at a distance of 10km from AM radio stations. ACKNOWLEDGEMENTS The authors would like to thank the National Institute of Standards and Technology (NIST) for funding this effort as part of the Technology Innovation Program (TIP) under Cooperative Agreement Number 70NANB9H9008. Additional support was provided by the Michigan Department of Transportation (MDOT), and the California Department of Transportation (Caltrans). We would like to thank our collaborators Dr. Jerome Lynch and Dr. Masahiro Kurata for providing the bridge acceleration data. REFERENCES [1] Lynch, J. P., "An overview of wireless structural health monitoring for civil structures," Philosophical Trans. of the Royal Society London, Series A (Mathematical, Physical and Engineering Sciences) 365(1851), (2007). [2] US Department of Transportation, Federal Highway Administration, "Our Nation's Highways: 2010," (2010). [3] Kurata, M., Lynch, J. P., Galchev, T., Flynn, M., Hipley, P., Jacob, V., van der Linden, G., Mortazawi, A., Najafi, K., Peterson, R. L., Li-Hong, S., Sylvester, D. and Thometz, E., "A two-tiered self-powered wireless monitoring system architecture for bridge health management," Proc. SPIE 7649, 76490K (2010). [4] Sohn, H., Farrar, C. R., Hemez, F. M., Shunk, D. D., Stinemates, S. W., Nadler, B. R., and Czarnecki, J. J., "Review of structural health monitoring literature from ," Report by Los Alamos National Laboratory, (2004). [5] Lynch, J. P., Law, K. H., Kiremidjian, A. S., Carryer, E., Farrar, C. R., Sohn, H., Allen, D. W., Nadler, B., and Wait, J. R., "Design and performance validation of a wireless sensing unit for structural monitoring applications," Structural Engineering and Mechanics 17(3-4), (2004). [6] Galchev, T., Kim, H., and Najafi, K., "Non-Resonant Bi-Stable Frequency Increased Power Generator for Low- Frequency Ambient Vibration," Digest of the 15th International Conference on Solid-State Sensors, Actuators, and Microsystems TRANSDUCERS' , (2009). [7] Galchev, T., "Energy scavenging from low frequency vibrations," PhD Dissertation, University of Michigan, Ann Arbor, (2010). [8] Galchev, T., McCullagh, J., Peterson, R. L., and Najafi, K., "A Vibration Harvesting System for Bridge Health Monitoring Applications," Digest of PowerMEMS 2010, (2010). [9] Kulah, H. and Najafi K., "Energy scavenging from low-frequency vibrations by using frequency up-conversion for wireless sensor applications," IEEE Sensors Journal 8(3), (2008). [10] Kulah, H. and Najafi K., "An electromagnetic micro power generator for low-frequency environmental vibrations," Digest of MEMS ' , (2004). [11] Balanis, C., [Antenna Theory: analysis and design, 2 nd ed.], Wiley, New York, (1997). [12] Snelling, E. C., [Soft Ferrites: properties and applications, 2 nd ed.], Butterworths, London, (1988). [13] Johnson, R. C., Jasik, H. and Knovel (Firm), [Antenna Engineering Handbook, 3 rd ed.], McGraw-Hill, New York, (1993) V. 1 (p.9 of 9) / Color: No / Format: Letter / Date: :31:54 PM