You are on page 1of 13

240

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

On Low-Frequency Electric Power Generation With PZT Ceramics


Stephen R. Platt, Shane Farritor, and Hani Haider
AbstractPiezoelectric materials have long been used as sensors and actuators, however their use as electrical generators is less established. A piezoelectric power generator has great potential for some remote applications such as in vivo sensors, embedded MEMS devices, and distributed networking. Such materials are capable of converting mechanical energy into electrical energy, but developing piezoelectric generators is challenging because of their poor source characteristics (high voltage, low current, high impedance) and relatively low power output. In the past these challenges have limited the development and application of piezoelectric generators, but the recent advent of extremely low power electrical and mechanical devices (e.g., MEMS) make such generators attractive. This paper presents a theoretical analysis of piezoelectric power generation that is veried with simulation and experimental results. Several important considerations in designing such generators are explored, including parameter identication, load matching, form factors, efciency, longevity, energy conversion and energy storage. Finally, an application of this analysis is presented where electrical energy is generated inside a prototype Total Knee Replacement (TKR) implant. Index TermsPiezoelectric materials, piezoelectricity, power generation, PZT ceramics.

I. INTRODUCTION AND RELATED WORK

ECHANICAL stresses applied to piezoelectric materials distort internal dipole moments and generate electrical potentials (voltages) in direct proportion to the applied forces. These same crystalline materials also lengthen or shorten in direct proportion to the magnitude and polarity of applied electric elds. Because of these properties, these materials have long been used as sensors and actuators. One of the earliest practical applications of piezoelectric materials was the development of the rst SONAR system in 1917 by Langevin who used quartz to transmit and receive ultrasonic waves [1]. In 1921, Cady rst proposed the use of quartz to control the resonant frequency of oscillators. Today, piezoelectric sensors (e.g., force, pressure, acceleration) and actuators (e.g., ultrasonic, micropositioning) are widely available. The same properties that make these materials useful for sensors can also be utilized to generate electricity. Such materials are capable of converting the mechanical energy of compression
Manuscript received August 4, 2003; revised October 11, 2003. This work was supported by the Christina M. Hixon Fund. S. R. Platt and S. Farritor are with the Department of Mechanical Engineering, University of Nebraska-Lincoln, Lincoln, NE 68588-0656 USA (e-mail: srp@unlserve.unl.edu; sfarritor@unl.edu). H. Haider is with the Department of Orthopeadic Surgery, University of Nebraska Medical Center, Omaha, NE 68198-1080 USA (e-mail: hhaider@unmc.edu). Digital Object Identier 10.1109/TMECH.2005.844704

into electrical energy, but developing piezoelectric generators is challenging because of their poor source characteristics (high voltage, low current, high impedance). This is especially true at low frequencies and relatively low power output. These challenges have limited the use of such generators primarily because the relatively small amount of available regulated electrical power has not been useful. The recent advent of extremely low power electrical and mechanical devices (e.g., microelectromechanical systems or MEMS) makes such generators attractive in several applications where remote power is required. Such applications are sometimes referred to as power scavenging and include in vivo sensors, embedded MEMS devices, and distributed networking. Several recent studies have investigated piezoelectric power generation. One study used lead zirconate titanate (PZT) wafers and exible, multilayer polyvinylidene uoride (PVDF) lms inside shoes to convert mechanical walking energy into usable electrical energy [2], [3]. This system has been proposed for mobile computing and was ultimately able to provide continuously 1.3 mW at 3 V when walking at a rate of 0.8 Hz. Other projects have used piezoelectric lms to extract electrical energy from mechanical vibration in machines to power MEMS devices [4]. This work extracted a very small amount of power ( W) from the vibration and no attempt was made to condition or store the energy. Similar work has extracted slightly W) from machine and building vibrations more energy ( [5]. Piezoelectric materials have also been studied to generate electricity from pressure variations in microhydraulic systems [6]. The power would presumably be used for MEMS but this work is still in the conceptual phase. Other work has used piezoelectric materials to convert kinetic energy into a spark to detonate an explosive projectile on impact [7]. Still other work has proposed using exible piezoelectric polymers for energy conversion in windmills [8], and to convert ow in oceans and rivers into electric power [9]. A recent medical application has proposed the use of piezoelectric materials to generate electricity to promote bone growth [10]. This work uses an implanted bone prosthesis containing a piezoelectric generator congured to deliver electric current to specic locations around the implant. This device uses unregulated (high voltage) energy and it is not clear if the technique has advanced beyond the conceptual phase. The above studies have all had some success in extracting electrical power from piezoelectric elements. However, many issues such as efciency, conditioning and storage have not been fully addressed. This paper presents an extensive theoretical analysis of piezoelectric power generation that is not presented

1083-4435/$20.00 2005 IEEE

PLATT et al.: ON LOW-FREQUENCY ELECTRIC POWER GENERATION WITH PZT CERAMICS

241

in the above studies. The analysis is veried with simulation and experimental results. Several important considerations in designing such generators are explored, including parameter identication, load matching, form factors, efciency, longevity, energy conversion, mechanical amplication, and energy storage. Finally, an application of this analysis is presented where electrical energy is generated inside an orthopedic Total Knee Replacement (TKR) implant [11]. The theoretical analysis is applicable to a wide range of force input waveforms with various frequencies. The experimental rekN) and low fresults are limited to relatively high forces ( quencies ( Hz.). This domain of experiments is chosen because of the practical limitations of experimental measurements and because it is most applicable to orthopedic implants and several other applications (e.g., [3], [5], [6], [9]). The experimental results were obtained by subjecting piezoelectric elements to various cyclic mechanical loads using a single-axis Mini-Bionix MTS 858 test machine. This machine is equipped with a PID controller and is capable of applying arbitrary force or displacement waveforms to the piezoelectric samples. Forces are measured using a calibrated load cell, and displacements are measured using an onboard linear variable differential transformer (LVDT). A PC running a custom Labview application communicating to a 12-bit National Instruments 6025E PCI DAQ card was used to generate force waveforms and to record all force, displacement, and voltage measurements. II. ELECTROMECHANICAL MODEL A. Piezoelectrics Piezoelectric materials convert energy between the mechanical and electrical domains. A linear isothermal model that describes the relationship between mechanical stress and strain and electric eld and electrical displacement (charge per unit area) was rst proposed by Voight [12] and further developed by Cady [13], Mason [14], and Jaffe and Berlincourt [15], among others. These constitutive relationships, and the techniques to measure piezoelectric material properties, were standardized in the IEEE Standard on Piezoelectricity [16]. The linear constitutive relationships are typically represented by the tensor equations

of an electric eld (short circuit electrical conditions), or the deection caused by an applied voltage in the absence of an applied force (stress-free mechanical conditions). In general, a material has three orthogonal coordinate axes along which the electric eld can be applied. For isotropic materials, the displacement will be parallel to the applied eld. However, for anisotropic materials like piezoelectric ceramics, this is not necessarily the case. Therefore, the dielectric permittivity is a 3 3 matrix and the electric eld and electric displacement are 3 1 vectors. Similarly, a longitudinal stress applied to a piezoelectric rod will not only cause the rod to become shorter and thicker, but it also tends to rotate about its longitudinal axis due to induced shear strains that do not exist in isotropic materials. To completely describe this behavior requires 21 independent stress and strain coefcients [17]. Thus, the stress and strain parameters in (1) are 6 1 matrices, and is a 6 6 compliance matrix. Many piezoelectric ceramics are manufactured or used in ways such that one or two of the piezoelectric strain coefcients dominate and the other interactions can be neglected. The relationships expressed in (1) then reduce to a pair of coupled equations that can be written as (2) where the subscripts and indicate the direction of the generated eld in response to an applied force. By convention, the axis labeled 3 is aligned with the poling axis. The current work uses a parallel compression generator where applied forces are . along the poling axis of the ceramic so that B. Electrical/Mechanical Model It is convenient to develop analytical models that can predict the response of these piezoelectric elements under different input conditions and for various material properties. A common approach for describing coupled electromechanical systems is to use lumped parameter models [18]. The principal assumption made in lumped parameter modeling is that the physical dimensions of the system are much smaller than the characteristic length scales of the governing electromagnetic, mechanical, or acoustic phenomena [18]. This allows the spatial and temporal variations of the system to be decoupled. The ideal piezoelectric coupling described by (2) can be rewritten as (3) where for simplicity the subscripts are not explicitly included, is the thickness of the material, and is the change in thickness. Assuming a sinusoidal steady-state operating condition, (3) can be differentiated with respect to time, transformed to the Laplace domain, and rearranged to yield (4) (5)

(1) is the electric displacewhere is the mechanical strain, is ment, is the mechanical stress, is the electric eld, the elastic compliance in a constant electric eld, is the piezois the dielectric constant under electric strain constant, and constant stress. The entire piezoelectric coupling is contained in the strain , the material is not piezoelectric and coefcients . If there is no coupling between the electric and mechanical elds. A larger value indicates stronger electromechanical coupling. The piezoelectric strain coefcient is the ratio of strain to applied eld or charge density to applied stress. Stated differently, measures the charge created by a given force in the absence

242

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

Fig. 1. Representation of piezoelectric coupling. Fig. 2. Lumped parameter model of mechanical system.

where velocity (m/s); current (C/s); short circuit compliance (m/N); free capacitance (C/V); volts (V); force (N). The relationships described by (4) and (5) can be represented by the circuit shown in Fig. 1 [18]. In this circuit, the coupling between the mechanical and electrical domains is modeled as an ideal transformer (derivation in the Appendix ). The transformer ratio can be thought of as the ratio of electrical output to mechanical input (V/N), and it can be determined from the piezoelectric strain constant and the physical compliance of the element [18] (6) It is important to recognize that the mechanical and electrical properties of piezoelectric materials change with load. For ), the charge generstress-free conditions (i.e., ated by an applied potential is given by (7) On the other hand, under blocked conditions (i.e., displacement constrained to be zero), the generated charge is given by

Using Newtons second law and the free body diagram shown in Fig. 2, the equation of motion for this system is given by (9) where external force (N); effective mass (kg); damping (N s m); stiffness compliance (N/m); mass displacement (m); mass velocity (m/s); mass acceleration (m s ). Taking the Laplace transform of (9) yields the system transfer function given by

(10) Because mechanical impedance is dened as the ratio of the force to the velocity of an element, these elements can be incorporated into the ideal representation of piezoelectric coupling as shown in Fig. 3(a). Reecting the mechanical elements of the lumped parameter model to the electrical side as shown in Fig. 3(b) yields equivalent electrical circuit element values given by ohms (11)

(8) kg where is the electromechanical coupling factor. Thus, the times smaller than the free capacblocked capacitance is itance. Similarly, the mechanical compliance under open circuit times smaller than the short cirelectrical conditions is cuit compliance. The circuit shown in Fig. 1 is an idealized representation of the piezoelectric effect, and it is most accurate only at low frequencies. Because this equivalent circuit neglects internal energy storage and loss mechanisms, several additional factors must be added to model the dynamic performance of piezoelectric elements at higher frequencies [18]. The mechanical properties of a real piezoelectric element can be modeled by a simple spring, mass, and damper system, shown in Fig. 2. The spring represents the ideal compliance, the mass is an energy storage element, and the damper accounts for internal mechanical losses. henries (12) farads Volts farads (13) (14) (15)

, shown in Fig. 3(b), accounts for the The extra term impedance of a load attached to the electrical output of the piezoelectric element. Here the load is shown as purely resistive, although that is not always necessarily the case. This model is still missing several factors such as dielectric loss. It is, in effect, a simplied linear representation of what is generally a much more complex nonlinear system. Piezoelectric ceramics have many different resonant modes, each of

PLATT et al.: ON LOW-FREQUENCY ELECTRIC POWER GENERATION WITH PZT CERAMICS

243

TABLE I PARAMETER FIT

Fig. 3.

Piezoelectric circuit model.

which would be characterized by a unique set of mechanical parameters. Hence, this model is likely valid only at frequencies below the rst resonant mode. Furthermore, piezoelectric material constants are not constants at all as they depend in a nonlinear fashion on conditions such as mechanical and electrical boundary conditions, and the frequency and amplitude of the excitation force prole. These effects must be considered when using this model to predict piezoelectric response under conditions signicantly different from those for which the model parameters are determined. III. EXPERIMENTAL FIT OF MODEL PARAMETERS All of the piezoelectric parameters in (11)(15) can in principle be determined from the various piezoelectric material constants and physical dimensions of the element. There are, however, a number of concerns with this approach. The impedance measurements used to determine piezoelectric parameters are usually small signal excitations at frequencies around 1 kHz. Therefore, these small-signal values may not be valid for highload low-frequency applications. Furthermore, because the parameters are known to change with external loading conditions, the values of the equivalent circuit elements can also be expected to depend on the electrical and mechanical boundary conditions. For these reasons, the values of the parameters used in this model are experimentally determined. The response of a passive electrical network such as the one depicted in Fig. 3(b) can be written in terms of the circuit elements and input signal. If the values of the circuit elements are unknown, they can be determined by measuring the system response to a known input. By modeling the behavior of a piezoelectric transducer as an analogous equivalent electrical circuit, the physical constants of the material can be determined by measuring its electrical response to a known input and using (11)(15) to relate the equivalent electrical parameters back to the physical mechanical properties. Under a sinusoidal input force prole, the amplitude of the voltage across the resistive load shown in Fig. 3(b) is given by

The amplitude of the input voltage , the parallel electrical impedance of the piezoelectric and the load , and the total circuit impedance . In terms of the circuit parameters in Fig. 3(b), the individual impedances are given by

(17) (18) (19) Equation (16) can be expanded and expressed in terms of the blocked electrical capacitance of the piezoelectric material ( ), the electrical equivalents ( , , and ) of the , , and ), and the transformer mechanical properties ( ratio . The unknown values of the four electrical elements and the transformer ratio are then determined by performing a nonlinear least squares t to the measured output voltages from a piezoelectric element under a known sinusoidal input force pro. le for various values of the load resistor, This approach was used to determine the parameters for a commercially available piezoelectric element. This rectangular (10 mm 10 mm 18 mm) element is composed of multiple ) wafers of lead zirconate titanate (PZT) ceramic (the ( advantage of using an element of this form is discussed in Section IV). The inset in Fig. 4 shows output voltage time history for a 1-Hz sinusoidal input force of amplitude 800 N. The average steady-state amplitude of the voltages measured across various load resistors are shown as squares. The curve represents the model t. The solid curve in Fig. 4 is the output predicted by (16) using the best t parameters listed in the rst column of Table I. The second column of Table I lists the values obtained for the lumped parameters calculated using the physical dimensions of the element and the piezoelectric constants for the PZT type 5 H material used in this element. There is excellent agreement between the two sets of values except for the mechanical damping coefcient which is not signicant at these frequencies.

(16)

244

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

Fig. 4. Voltage across the resistive load and the model t.

This electromechanical model can be expected to provide some qualitative insight into the behavior of the piezoelectric elements under conditions in the neighborhood of that for which the model parameters are determined. Once veried, the model can then be used with other simulation tools to help optimize power conditioning and storage strategies without the need to physically incorporate the piezoelectric elements during the early design stages. IV. RAW POWER A. Time History and Impedance Matching Plot The instantaneous raw electrical power dissipated in the . The resistive load shown in Fig. 3(b) is given by total power generated depends on the resistive load and applied stress. The main graph in Fig. 5 shows the average power generated for a 1-Hz 800-N amplitude sine wave force input as a function of the resistive load. The squares are the experimentally determined values, and the solid curve is the output predicted by the electromechanical model. The inset in Fig. 5 shows representative time-series results for the instantaneous voltage (middle plot) and power (bottom plot) delivered to a 14.8-k resistive load. The horizontal line is the average power delivered to the load ( 1.08 mW.) These results show the importance of matching the load to maximize the power generated. At low values of resistance no voltage is produced (short circuit) and no power is generated. At high resistances (open circuit) no current ows and no power is generated.

B. Form Factors: Stacks and Transformers As described in the mathematical model, the size and shape of the piezoelectric element affect the amount of power generated. First, for a given load, the mechanical stress (and strain) is related to the cross sectional area. The voltage produced is related to the stress, and the power (for a resistive load) is related to the square of the voltage. Second, for a given stress the charge displacement is related to the volume of the element [ and in (7)]. The mechanical design of a piezoelectric element also has important implications on the electrical source characteristics. Consider the two commercially available forms of PZT piezoelectric ceramic shown in Fig. 6. The element on the left is a homogenous monolithic cylinder 1.0 cm in diameter and 2.0 cm long (Sensor Technology). This form has a low capacitance ( 47 pF) and produces a very high open circuit voltage ( 10 000 V). The second element is manufactured using a proprietary technique by stacking multiple ( 145) PZT wafers. Each wafer has a pair of electrodes and they are assembled mechanically in series but electrically in parallel (Piezo Systems). The wafers are polarized along their thickness so that they exhibit a piezoelectric effect only in this direction. These stacks are typically used as actuators, and the multiple layers have the effect of increasing the effective capacitance (110 F) and decreasing the peak open circuit voltage ( 30 V). The stack is 1.0 cm square and 1.8 cm long. The effect of this mechanical design is shown in Fig. 7. The curve on the left shows the experimental and simulation results for the stack (similar to Fig. 6, right). The curve on the right

PLATT et al.: ON LOW-FREQUENCY ELECTRIC POWER GENERATION WITH PZT CERAMICS

245

Fig. 5. Power as a function of resistive load: 1-Hz 800-N sine wave.

Fig. 6.

Form factors: monolithic and stacks.

shows simulation results for a monolithic element of exactly the same material, volume, and cross-sectional area (similar to Fig. 6, left). Only simulation results are shown because a monolithic cylinder of the same material, volume and cross sectional area was not commercially available. However, the general characteristics of cylindrical monolithic elements were experimentally veried using the Sensor Technology PZT type 4 element in a series of experiments and simulations as described in [11]. It is important to notice that under the same force loading conditions, both elements produce the same power into matched resistive loads ( 1.1 mW), although the matched load is much lower ( k ) for the stack than the cylinder ( G ).

Fig. 7. Average power delivered to R for the stack (left) and the monolithic element (right).

Stacking several elements mechanically in series (sharing the same stress; strain is additive) and electrically in parallel

246

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

(sharing the same voltage; current is additive) has the effect of stepping down peak voltages and stepping up the effective capacitance compared to the monolithic element. This is in effect a mechanical transformer. A similar mechanical transformer effect could be obtained by using other mechanical geometries such as rows of piezoelectric lm on the surface of a beam in bending. The same electrical effect can be obtained by using an electrical transformer to reduce the output voltage of the monolithic element (although, because of the high voltages, arcing would be a problem). These results show that: 1) it is critical to match the circuit load with the type of generating material, and 2) stacked and monolithic elements of the same geometry produce the same power into matched loads. These results also indicate that the stacked element is preferred because of the lower matching electrical load required. The stacked element reduces the output voltage to a more manageable level, and the higher capacitance of this form is important for energy transfer (discussed in Section VI). For these reasons, the piezoelectric stacks are used in the remainder of this work. V. EFFICIENCY OF POWER CONVERSION The network depicted in Fig. 3(b) can be used to explore the overall efciency of converting mechanical power into electrical power. Because the voltage and current will be in phase across the load resistor, the average power delivered to the load is given by (20) The most conservative estimate of the efciency for power transfer to the load is given by the ratio of the average output power to the magnitude of the total apparent input power (21) where the root mean square (rms) amplitude of the input current is given by (22) The surface of Fig. 8 shows numerical results calculated according to (21) for the overall electromechanical efciency for the raw power produced by the piezoelectric stack. Again, the surface represents the simulation results and the squares and associated lines represent individual experiments. The input force prole was sinusoidal with 800-N amplitude, and the efciency was calculated for a range of frequencies and load resistors, using the best t model parameters as listed in Table I. The results of this evaluation indicate that signicant efciencies are obtained only along the narrow ridge of load resistor values for which the source and load impedances match (as in Section V). As the input frequency increases, the maximum efciency occurs at smaller load resistor values. This is the behavior expected for an essentially capacitive device; as the excitation , frequency increases, the source impedance, will decrease.

Fig. 8. Efciency as a function of frequency and resistive load.

There is also a trend toward higher efciency as the excitation frequency increases. The efciency peaks as would be expected at the resonant frequency of about 34 kHz. Note, however, that any conclusions drawn from these trends must be qualied by the restrictions and limitations of the model used (as discussed in Section II). In particular, the behavior of piezoelectric ceramics is generally nonlinear, and this is especially true at and around the resonant frequency. Furthermore, the electromechanical model parameters are not constant as the input frequency increases. Given the restrictions at higher frequencies, the validity of the model was veried by comparing the predicted efciency to that actually observed, for lower frequencies. For a sinusoidal , the total apparent input input force with rms amplitude , can be calculated directly as mechanical power, (23) is the rms amplitude of the velocity, determined where from the measured displacement of the element. The output power and efciency are calculated as usual using (20) and (21). Measurements were made for a range of input frequencies and resistive loads. Fig. 9 shows the predicted and actual efciencies for 1, 5, and 10 Hz, and the solid lines and circles shown on Fig. 8 show the experimentally determined efciencies superimposed on the complete numerically calculated dataset. The agreement between the model and experimentally measured data points is excellent and serves to further validate the electromechanical model at these frequencies. VI. ENERGY CONVERSION AND STORAGE A. Introduction Efciently transforming the raw power from a piezoelectric generator into useful power is not straightforward. The results presented in Sections V and VI show that the low-frequency properties of this type of generator are characterized by highly capacitive source impedance, and low-currenthigh-voltage ac (bipolar with large ripple) electrical output. This is clearly not appropriate for electrical components and sensors without signicant signal conditioning. This section describes some important issues in conversion and storage of the generated energy.

PLATT et al.: ON LOW-FREQUENCY ELECTRIC POWER GENERATION WITH PZT CERAMICS

247

Fig. 10.

Energy transfer to a storage capacitor.

Fig. 9.

Electromechanical efciency.

B. Source Capacitance One method to smooth electrical signals and store energy is to include a capacitor (and rectier to convert the bipolar piezoelectric output into a unipolar signal) in parallel with the resisis high and tive load. The capacitor stores energy when is low. Such a capacitor might also releases energy when represent other forms of storage such as a battery that is charged by the piezoelectric generator. A difculty with this approach is that the capacitance of the piezoelectric ( ) makes the direct transfer of energy into a storage capacitor ( ) inefcient. Consider the simple model shown in Fig. 10. Here the piezoelectric is charged (compressed) to create an and an initial charge . The initial system initial potential energy (assuming no initial energy on the storage capacitor ) is given by (24) After the switch is closed both capacitors will have the same voltage and the charge will distribute according to the magnitude of the capacitors. The nal stored energy, assuming no losses, is given by

Because piezoelectric materials at low frequencies are essentially purely capacitive devices, energy transfer losses can be minimized by reducing the size of the storage capacitor. However, in this case, no energy is transferred. Maximum energy is and the maximum value is . transferred when This is a very simplied model of energy transfer from a piezoelectric generator, but it exemplies the inefciency inherent in the process. Other studies have addressed this problem by allowing the piezoelectric to build to a peak voltage before ( ) and, therefore, closing the switch [3]. This increases increases energy transferred to the storage capacitor. However, efciency is still determined by matching capacitances. C. RC Loading To quantify the inefciencies associated with a direct energy transfer approach, three piezoelectric stacks were connected to in parallel with a rea load consisting of a storage capacitor . Full-wave bridge rectiers were also inserted sistive load between the piezoelectric elements and the load to convert the bipolar piezoelectric output into a unipolar signal. The piezoelectric stack combination was subjected to a 2600-N amplitude input force, and the power delivered to a resistive load by the generator was determined for a wide range of storage capacitors and electrical loads. Fig. 11 shows the design space with simulation (mesh) and laboratory results for cross sections of constant load rek ) and constant storage capacitance sistance ( F). ( For low values of the storage capacitor, the maximum power delivered to a matched resistive load approaches the raw power results (Section IV) with the difference being due to losses in in effect reduce the system to the rectiers. Small values of a rectied version of the raw power experiment and maximize does not store much the power generated. However, a small energy and will have little effect on the level of output ripple in the time domain. Increasing the value of the storage capacitor reduces the amount of ripple in the time domain but also reduces the energy transfer efciency. Using a 10 F capacitor reduces the efciency of this circuit to about 40% compared to the total raw power delivered by three piezoelectric elements to a purely resistive load. Furthermore, examination of the time-series data shows that the voltage across the load still has signicant ( 100%) ripple. The level of ripple is only signicantly reduced for larger values of , but then energy transfer efciency rapidly approaches zero. The surface in Fig. 11 shows that maximum energy is delivered to the load resistor when the storage capacitor is small and

(25) Therefore, the nal system energy is reduced. These losses can occur in the arcing of the switch or as resistive losses in half the energy is lost and the energy the wires. If transferred to the storage capacitor is (26)

248

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

Fig. 11.

Power generated versus R and C .

the load resistor matches the impedance of the source. Any combination of load resistor and storage capacitor chosen to minimize the level of ripple will move the operating point into regions of low energy transfer. These results illustrate some of the difculties associated with low frequency piezoelectric power generators. While direct transfer of energy from a piezoelectric element is possible, especially for high capacitance stacks like the ones used in this work, the highest efciencies will only be realized with power conditioning systems optimized to address the capacitive nature of piezoelectric elements. D. Electrical Conditioning The above analysis is important because the efcient use of a storage capacitor (or battery) is important to supply energy when the piezoelectric element has low potential. However, a capacitor alone may not provide sufcient signal conditioning. Therefore, a combination of a capacitive storage device and a voltage regulator may be required. Voltage regulators are well-established devices used to create an output voltage that is constant and independent of the load and supply variations (e.g., [19]). They appear to the source as constant-current loads and are very effective at reducing ripple by eliminating voltage peaks. The proper combination of storage capacitor and voltage regulator, as shown in the application in Section IX, can be used to condition the piezoelectric output to produce a more useable power supply. VII. LONGEVITY It is well known that repeated mechanical and electrical cycling of piezoelectric ceramics results in a progressive degradation in performance as a function of time [20][22]. During typical operation, a piezoelectric device experiences repeated reversals of the electric drive eld and/or mechanical load, along with cyclic variations in temperature. These effects contribute to performance degradation through mechanisms such as the loss

of mobility in domain walls, depoling, the formation of microcracks, and the failure of the electrode-ceramic interface [23]. The degradation in the performance of piezoelectric ceramics is often characterized in terms of logarithmic functions of time. The actual changes in performance depend on application specic details such as the amplitude, frequency, and duration of the applied electrical and mechanical loads, as well as the particular composition of the PZT material [24]. Limited data currently exist on the performance characteristics of piezoelectric ceramics subject to the low frequency, high mechanical loading conditions that are likely to exist in many potential applications for piezoelectric power generation. In an attempt to characterize the long-term performance of piezoelectric elements under these conditions, the raw electrical power delivered to a resistive load was monitored for approximately 1.5 million cycles. A single Piezo Systems piezoelectric stack was subjected to a 440-N amplitude composite input force prole containing several low ( 20 Hz) frequency components while its electrical output was monitored across a 29.5-k resistive load. The applied force prole is the expected loading in the application discussed in Section IX. The force prole was periodic at a frequency of 1 Hz so the number of cycles corresponds directly to time measured in seconds. The average power delivered to the resistive load, shown in Fig. 12, was calculated every 30 s for the rst 60 min, and then every 30 min thereafter. Several features are immediately apparent. Little or no degradation in electrical output is obvious during approximately the cycles. Subsequently, the output decreases approxirst mately logarithmically per decade of time. It was also observed, but not shown in this plot, that a pause in the experiment (30 min or more) can cause signicant recovery of the piezoelectric generator [11]. This experiment suggests that: 1) this piezoelectric generator would create useful power for millions of cycles, and 2) the exact amount of decay in performance is application dependent. The results shown in Fig. 12 are not generalthe longevity of the generator depends on several application specic factors

PLATT et al.: ON LOW-FREQUENCY ELECTRIC POWER GENERATION WITH PZT CERAMICS

249

Fig. 12.

Power as a function of time.

Fig. 13.

Mechanical amplication examples.

(e.g., loading stress, force prole, excitation frequency, temperature variations). The results do verify that the logarithmic progressive degradation of piezoelectric materials is also seen when these materials are used as power generators. VIII. MECHANICAL AMPLIFICATION The electrical energy stored by a xed volume of piezoelectric material can be written using (2) as (27) where is the volume of the element and is the square of the applied stress. Equation (27) implies that more energy is available from a larger element. This is similar to stating that a larger battery has more energy. However, (27) also dictates that the available energy is proportional to the square of the applied stress. The implication is that not only is material selection important in maximizing available power, the details of form factor and force application are as well. More power may be available by choosing a force amplication scheme applied to fewer or smaller piezoelectric elements, rather than attempting to maximize the volume of material. Fig. 13 shows two techniques that can be used to increase the power generation for a given load prole. In the left gure a locking mechanism is used to amplify the axial force in each link and thereby increase the strain in the piezoelectric. In the gure on the right, the relatively low resonance of a mechanical cantilever can be used to increase the strain delivered to a piezoelectric lm on the surface of the cantilever (similar to [5]).
Fig. 14. Self-powered TKR model and test setup. (a) Components of the implant model. (b) Test setup.

It should be noted that while the power generation is increased with higher strain, the longevity of the piezoelectric material is decreased. IX. POWER GENERATION IN TOTAL KNEE REPLACEMENT IMPLANTS The principles of piezoelectric power generation described above are now applied to the design of a self-powered Total Knee Replacement (TKR) implant. Sensors encapsulated within implants could provide in vivo diagnostic capabilities such as the monitoring of implant duty (i.e., walking) cycle, detecting abnormally asymmetric or high forces, sensing misalignment and early loosening, and early detection of wear. Early diagnosis of abnormalities is critical to minimize patient harm. Because implants are expected to last more than 20 years, the lack of a long-term self-contained power supply is a major limitation to implant sensors. A simplied design of a self-powered instrumented implant is shown in Fig. 14. This design is not intended to fully address the complex details of implant design, but rather to demonstrate the feasibility of piezoelectric power generation. Many additional

250

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

Fig. 15.

Electrical design.

constraints (e.g., wear, bio-fouling, off axis loading) must be considered before clinical implementation is possible. A typical xed bearing TKR implant has three major components. The femoral component is a highly polished hard surface that is attached to the femur. The tibial tray is attached to the tibial bone and supports a low friction polyethylene bearing surface. The tray has a stem that extends into the bone to help support moments and nonaxial forces. Relative motion between the femoral and bearing surfaces allows the knee to function. Fig. 14(a) shows a tibial tray, similar to traditional TKR designs, that has been modied by making it deeper and including three stems to accommodate three piezoelectric stacks. Force is applied by the femoral component to the bearing surface and a distribution plate is used to transfer the load to the three elements. The elements are approximately located so an axial force is evenly distributed in each element. However, in a real implant, the forces are multi-axial and the contact points between the femoral component and the bearing move with any combination of exion, anteriorposterior motion, and axial rotation. Therefore, the forces would not always be equally distributed. The electrical design discussed below will accommodate for various force distributions. Fig. 14(b) shows the implant model mounted on an anatomically correct synthetic tibia and femur (Sawbones). Output voltage was regulated with an active voltage regulator, as shown in the electrical circuit in Fig. 15. The output of each of the three piezoelectric elements was rectied using full-wave bridge rectiers. This unipolar output was then ltered and stored with a 10- F storage capacitor. This capacitor is chosen to approximately match the effective capacitance of the three elements. A simple low-power MAX 666 linear regulator was used (more complex switching regulators will likely be incorporated in future work). The output of the regulator was used to power a PIC 16LF872 microprocessor. This is a basic low-power processor with both digital I/O and analog inputs.

Tests were performed where the standard 2600-N amplitude force prole [24] for the axial force in the knee during normal walking was applied to the TKR model. As much as 4 mW of raw electrical power and 0.85 mW of regulated power was generated in this application. The overall electrical efciency in delivering regulated power was approximately 19%. Additional tests were performed with the amplitude of the applied force reduced by 50%. The top plot in Fig. 16 shows three cycles of the applied ISO axial force prole for normal walking [25]. The middle plot shows the regulated output voltage from the MAX 666. The bottom plot shows the output status of an LED attached to the microprocessor. The PIC microprocessor was programmed to turn the LED on for a xed period during the applied force cycle. This output then represents implant duty. The processor could perform many other operations such as measuring peak forces and/or force distributions by monitoring the rectied voltage of the piezoelectric elements. The processor could also store the necessary data in onboard nonvolatile memory until it could be downloaded during the patients next clinical visit. Even under this reduced force condition, approximately 225 W of continuous regulated power is available. This is more than adequate to satisfy the approximately 50- W power requirement of the PIC 16LF872. The experimental results presented here use the energy as it is generated. The generator could also be used, based on the analysis presented in Section VI, to charge a battery so that energy is available when the patient is not walking. X. CONCLUSION Piezoelectric power generators have great potential for some remote applications such as in vivo sensors, embedded MEMS devices, and distributed networking. However, developing a piezoelectric generator is challenging because of their poor source characteristics (high voltage, low current, high

PLATT et al.: ON LOW-FREQUENCY ELECTRIC POWER GENERATION WITH PZT CERAMICS

251

(28) can be solved for the velocity (31) Substituting this result into (29) and recognizing that (32) yields for the current (33) which is the same as (5) if (34) Substituting (33) into (31) yields the remaining piezoelectric coupling equation, (4). ACKNOWLEDGMENT The experimental work was performed at the Biomechanics Laboratory in the Department of Orthopedic Surgery at the University of Nebraska Medical Center. The authors would like to acknowledge the support of Dr. K. Garvin, M.D., Chairman of the Department of Orthopedic Surgery, UNMC. REFERENCES
Fig. 16. Regulated power output and microprocessor output. [1] K. F. Graff, A history of ultrasonics, in Phys. Acoust.. New York: Academic, 1981, vol. 15, ch. 1. [2] J. Kymissis, C. Kendall, J. J. Paradiso, and N. Gershenfeld, Parasitic power harvesting in shoes, in Proc. 2nd IEEE Int. Conf. Wearable Computing, Los Alamitos, CA, Aug. 1998, pp. 132139. [3] N. S. Shenck and J. A. Paradiso, Energy scavenging with shoe-mounted piezoelectrics, IEEE Micro, vol. 21, no. 3, pp. 3042, May-Jun. 2001. [4] P. Glynne-Jones, S. P. Beeby, and N. M. White, Towards a piezoelectric vibration-powered microgenerator, IEE Proc. Sci. Meas. Technol., vol. 148, no. 2, pp. 6872, 2001. [5] S. Roundy, The power of good vibrations, Lab Notes-Research from the College of Engineering, University of California, Berkeley, vol. 2, no. 1, Jan. 2002. [6] N. W. Hagood IV et al., Development of micro-hydraulic transducer technology, in Proc. 10th Int. Conf. Adaptive Structures and Technologies, Paris, France, Oct. 1999, pp. 7181. [7] T. G. Engel, Energy conversion and high power pulse production using miniature piezoelectric compressors, IEEE Trans. Plasma Sci., vol. 28, no. 5, pp. 13381340, Oct. 2000. [8] V. Hugo Schmidt, Piezoelectric energy conversion in windmills, in Proc. Ultrasonics Symp., 1992, pp. 897904. [9] G. W. Taylor, J. R. Burns, S. M. Kammann, W. B. Powers, and T. R. Welsh, The energy harvesting eel: A small subsurface ocean/river power generator, IEEE J. Ocean. Eng., vol. 26, no. 4, pp. 539547, Oct. 2001. [10] C. S. McDowell, Implanted bone stimulator and prosthesis system and method of enhancing bone growth, U.S. Patent 6,143,035, Nov. 7, 2000. [11] S. R. Platt, Electric power generation within orthopaedic implants using piezoelectric ceramics, Masters thesis, Dept. Mech. Eng. Univ. Nebraska-Lincoln, 2003. [12] V. W. Voigt, Lehrbuch der Kristallphysik. Liepzig, Berlin, Germany: B. G. Teubner, 1910. [13] W. G. Cady, Piezoelectricity: An Introduction to the Theory and Applications of Electromechanical Pheonomena in Crystals. New York: Dover, 1964, vol. 1 & 2. [14] W. P. Mason, Piezoelectric Crystals and their Application to Ultrasonics. New York: Van Nostrand, 1950. [15] H. Jaffe and D. A. Berlincourt, Piezoelectric transducer materials, Proc. IEEE, vol. 53, pp. 13721386, 1965. [16] IEEE/ANSI 176 IEEE Standard on Piezoelectricity, 1987. [17] V. E. Bottom, Introduction to Quartz Crystal Unit Design. New York: Van Nostrand Reinhold, 1982.

impedance) and relatively low power output. The recent advent of extremely low power electrical and mechanical devices (e.g., MEMS) makes such generators attractive. This paper presents a theoretical analysis of power generation with PZT ceramics that is veried with simulation and experimental results. Several important considerations in designing such generators are explored, including parameter identication, load matching, form factors, efciency, longevity, energy conversion, and energy storage. Finally, an application of this analysis is presented where electrical energy is generated inside a prototype Total Knee Replacement (TKR) implant. APPENDIX That the circuit shown in Fig. 1 accurately represents the piezoelectric coupling equations can be seen as follows. Applying Kirchoffs current law to the node on the electrical side of the circuit yields (28) Applying the mechanical analog of Kirchoffs voltage law to the mesh on the mechanical side of the circuit yields (29) Using the fact that (30)

252

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 10, NO. 2, APRIL 2005

[18] M. Rossi, Acoustics and Electroacoustics. Norwood, MA: Artech House, 1988. [19] P. Horowitz and W. Hill, The Art of Electronics, 2nd ed. Cambridge, U.K.: Cambridge Univ. Press, 1989. [20] H. H. A. Krueger and D. Berlincourt, J. Acoust. Soc. Am., vol. 33, pp. 13391344, 1961. [21] M. D. Hill, G. S. White, and C.-S. Hwang, Cyclic damage in lead zirconate titanate, J. Am. Ceram. Soc., vol. 79, no. 7, pp. 19151920, 1996. [22] M. G. Cain, M. Stewart, and M. G. Gee, Degradation of piezoelectric materials, National Physical Laboratory Management Ltd., Teddington, Middlesex, U.K., NPL Rep. SMMT (A) 148, 1999. [23] F. Lowrie, M. Cain, and M. Stewart, Time dependent behavior of piezoelectric materials, National Physical Laboratory Management Ltd., Teddington, Middlesex, U.K., NPL Rep. SMMT (A) 151, 1999. [24] G. Yang, S.-F. Liu, W. Ren, and B. K. Mukherjee, Uniaxial stress dependence of the piezoelectric properties of lead zirconate titanate ceramics, in Active Materials: Behavior and Mechanics. Bellingham, WA: SPIE, 2000, vol. 3992, SPIE Proceedings, pp. 103113. [25] Implants for SurgeryWear of Total Knee-Joint ProsthesesPart 1: Loading and Displacement Parameters for Wear-Testing Machines with Load Control and Corresponding Environmental Conditions for Test, ISO 14243-1:2002(E), Mar. 15, 2002.

Shane Farritor received the B.S. degree in mechanical engineering from the University of Nebraska-Lincoln in 1992, and the M.S. and Ph.D. degrees in mechanical engineering from the Massachusetts Institute of Technology, Cambridge, in 1998. He is currently an Assistant Professor in the Department of Mechanical Engineering, University of Nebraska-Lincoln, and holds courtesy appointments in both the Department of Surgery and the Department of Orthopaedic Surgery at the University of Nebraska Medical Center, Omaha. His research interests include space robotics, surgical robotics, biomedical sensors, and robotics for highway safety. Dr. Farritor serves on both the AIAA Space Robotics and Automation technical committee and the ASME Dynamic Systems and Control Robotics Panel.

Stephen R. Platt received the B.A. degree in physics and astronomy from Williams College, Williamstown, MA, in 1983, the M.S. and Ph.D. degrees in astronomy and astrophysics from the University of Chicago, Chicago, IL, in 1991, and the M.S. degree in mechanical engineering from the University of Nebraska-Lincoln in 2003. He is currently a Research Assistant Professor in the Department of Mechanical Engineering, University of Nebraska-Lincoln. His interests include biomedical sensors, surgical robotics, and millimeter-wave detector system for observational astrophysics applications.

Haini Haider was born in Baghdad, Iraq, in 1960. He received the B.Eng. and Ph.D. degrees in mechanical engineering from the University of Shefeld, Shefeld, U.K., in 1983 and 1990, respectively. He was on the faculty of the University of Shefeld from 1988 to 1996, where he became a Lecturer in 1992. His research then encompassed uid dynamics, mechatronics, robotics, and information technology. In 1997, he joined the faculty of University College London at the Centre of Biomedical Engineering in Stanmore. He was the principal mechanical and software engineer on the team of Dr. Peter Walker that produced the Instron-Stanmore Knee Simulator and the International Standards Organization (ISO) method for simulation and wear testing of knee replacement systems. He joined the faculty of the University of Nebraska Medical Center, Omaha, in March 2000 as an Associate Professor of Biomedical Engineering at the Department of Orthopaedic Surgery and Rehabilitation. He is the author of over 65 publications and international conference presentations in engineering and orthopedics biomechanics research. His current research interests involve innovative implant design, testing, computer-aided simulation, preoperative planning, and image-guided and robotic surgery. Dr. Haider is a Member of the Engineering Council (U.K.), Member of the Institution of Mechanical Engineers (I.Mech.E) (U.K.), Member of the American Society of Testing and Materials (ASTM) and co-chairs its committee for Knee Implant Wear Testing, Chair of the International Standards Organization Review Workgroup of ISO 14243-1 and 14243-2, Member of the International Society of Technology in Arthroplasty (ISTA), and Member of the Orthopaedic Research Society (ORS). He is the winner of academic prizes including the Klinger International Research Prize, Siebersdorf, Austria.

You might also like