Gregory G. Weiss Lockheed/Martin Astronautics, P.O. Box 179, MS T500, Denver, Colorado 80201

Transcription

1 Body waves recorded inside an elastic half-space by an embedded, wideband velocity sensor Steven D. Glaser Department of Civil and Environmental Engineering, University of California, Berkeley, California Gregory G. Weiss Lockheed/Martin Astronautics, P.O. Box 179, MS T500, Denver, Colorado Lane R. Johnson Department of Geology, University of California, Berkeley, California Received 10 October 1997; accepted for publication 18 May 1998 This paper presents a unique embeddible acoustic emission sensor. Comparison with theoretically calculated waveforms for the embedded sensor, surface step force Lamb s problem prove the sensor to be an accurate transducer of particle velocity, with sensitivity of 2.34 V output per mm/s. Calibration as a surface sensor by the National Institute of Standards and Technology NIST show the sensor to be an accurate transducer of surface displacement with a sensitivity of 2.8 V/nm. The paper presents details about the design and construction of the sensor as well as calibration and verification. Unique design elements include the use of a lead-alloy backing masses, soft elastomer shear-spring isolation and mounting, and embedment. The sensor is based on the NIST conical lead zirconate-titanate PZT element and has a finished length of 38 mm and a diameter of 16 mm. The sensor is robust enough to work under 1 MPa of brine pressure Acoustical Society of America. S X PACS numbers: Ar, Le, Fx SLE INTRODUCTION Traditionally, acoustic emission AE studies are conducted by placing an array of sensors on the surface of the specimen being tested. Unfortunately the physical constraints of many practical testing geometries preclude surface placement of sensors. A simple solution to this problem is to place the sensors inside the body to be monitored, but to date no such sensors have been available. This paper introduces a unique, embeddible, high-fidelity particle velocity transducer, including design parameters and National Institute of Standards and Technology NIST calibration. The design yields sensitivity equal to commercial resonant devices while maintaining the high fidelity of the NIST displacement sensors. The embeddible high-fidelity sensor discussed here for the first time allows experimental verification of Lamb s problem of body motions due to a surface excitation. Validation of the performance of the sensors is provided by a parametric study of recorded and theoretical waveforms for Lamb s problem, buried transducer, surface source. Using embedded sensors has several advantages over surface mounted sensors, including: i mode conversion is avoided so the measured waveforms are simpler and easier to interpret; ii a greater length of signal can be recorded before being contaminated by surface reflections/mode conversions; iii the sensor s can be located closer to the area of crack growth; iv embedded sensors may be less susceptible to accidental movement or damage; v embedded sensors are the only option when free surfaces are not accessible Glaser, Most acoustic emission studies use resonant sensors because they provide greater sensitivity near the resonance frequency. Unfortunately, resonant devices lack the bandwidth needed to analyze incoming waveforms with the rigor with which seismologists study earthquake waveforms e.g., Aki and Richards, Resonant sensors provide a reasonably accurate estimate of the AE time of arrival, but beyond the direction of first motion the received signal is more a function of the sensor than of the kinematics at the crack tip e.g., Pao, 1978; Hamstad and Fortunko, An acoustic emission study using resonant sensors is empirical, relies on correlations, and suffers from a very thin causal science basis Eitzen et al., By using a wideband sensor having a flat frequency response over several octaves, the recorded signal is equivalent to the actual kinetics at the receiver location, enabling researchers to determine not only the time of first arrival, but also the arrival times of the P wave, S wave, and boundary reflections, as well as their relative magnitudes e.g., Glaser and Nelson, 1992b; Kishi, It has been shown that by using a high-fidelity sensor, the inverse problem of determining the source function from remote measurements can be achieved e.g., Eisenblatter, 1980; Eitzen et al., 1981; Kim and Sachse, Full waveform signals also allow the use of forward modeling to evaluate source kinematics e.g., Aizawa et al., I. BACKGROUND AND THEORY A. Lamb s problem Wave propagation problems involving an elastic halfspace have come to be known as Lamb s problems due to the early work in the field by Horace Lamb Lamb looked 1404 J. Acoust. Soc. Am. 104 (3), Pt. 1, September /98/104(3)/1404/9/$ Acoustical Society of America 1404

2 TABLE I. Mechanical properties of the dolostone test block. Young s modulus, E GPa Shear modulus, G GPa Bulk modulus, K GPa Poisson s ratio, Density, (kg/m 3 ) v p v s mm/ s Q p mm/ s Q s at surface response due to a surface excitation by integrating Rayleigh s 1887 discovery of the surface wave. Both researchers were attempting to understand surface response to earthquake excitation. Lamb noted that the P- and S-wave history minor tremors was a function of the time derivative of the source function while the surface wave major tremor history was a direct function of the source kinematics Lamb, The first partial solution to the more complicated halfspace problem, internal motions due to surface excitation, was published by Miller and Pursey Miller and Pursey derived integral representations of the internal motions due to simple harmonic motions normal to the surface of the half-space. Pekeris provided a full solution for the motion of the half-space surface due to a buried Heaviside pulse in terms of a single integral Pekeris and Lifson, Five years later Cherry 1962 published a solution for body and surface wave propagation due to a horizontal motion on the surface of the half-space. Johnson 1974 presented a complete solution for the three-dimensional Lamb s problem in the form of a Green s function and spacial derivatives for an elastic, uniform halfspace. Ohtsu and Ono 1984 published a computer program simplifying Johnson s numerical procedure for the surface receiver interior source case. Using ray theory, Pao and Gajewski 1977 provided general solutions for transient waves generated by a multitude of sources for layered solids, with more detailed solutions derived by Mal e.g., Mal et al., Application of these solutions to decoding acoustic emission source events include work by Wadley e.g., Wadley and Hill, 1983 and Michaels Michaels et al., constructed, with further details in Weiss and Glaser A. Sensor element The embedded sensor uses a 2.5-mm-tall truncated cone of PZT-5a a lead-zirconate-titanate composition, with an aperture diameter of 1.5 mm and a base diameter of 6.5 mm. This geometric design of piezoelectric cone was developed by the National Institute of Standards and Technology NIST for use in wideband acoustic emission sensors e.g., Eitzen et al., 1981; Proctor, 1982a, b. The conical design eliminates the aperture effect by keeping the contact area small, is sensitive to only one parameter motion in line with the cone axis, and reduces the degeneracies of the normal modes of the usual piezoelectric disk element Proctor, 1982b; Greenspan, The element has a high piezoelectric coupling constant and a mechanical stiffness estimated to be N/m at a minimum Hamstad and Fortunko, B. Electronics system Only a small fraction of the energy released from an acoustic emission event ever reaches a sensor. Furthermore, a piezoelectric sensor converts only a fraction of the received energy into an electrical signal. Unless this signal is amplified at the crystal, it will be corrupted by electromagnetic noise and capacitive loading from the cable between the sen- B. Solution of Lamb s problem Using the viscoelastic propagator method introduced by Kennett 1983, synthetic seismograms were generated for a uniform half-space having the elastic properties listed in Table I. The attenuation properties of the rock half-space were modeled with the quality factors Q p Q s 50, and the constant Q model of Kjartansson 1979 was used for the dispersion relationship. These synthetic seismograms are an appropriate solution for our problem: a unit step function in force acting on the free surface, a receiver that records vertical motion at a depth of 152 mm within the half-space, and at a range of horizontal offsets from the source position. The calculations were performed for a buried source and surface receiver and then reciprocity was used to obtain the results for the experimental geometry. II. DESIGN OF THE EMBEDDED HIGH-FIDELITY ACOUSTIC EMISSION SENSOR A cutaway schematic of the sensor as built is shown in Fig. 1. The following sections explain how the sensor is FIG. 1. Cutaway schematic drawing of the embeddible sensor J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1405

3 Previous research Proctor, 1982a, 1986 has shown that the geometry of the backing mass is the single most important element of wideband acoustic emission sensor design. In general, a larger backing mass will have fewer resonances and will give a more satisfactory performance at lower frequencies. A unique problem was presented since the embedded sensor was not supposed to affect the passing wavetrain. Thus the backing mass has been carefully designed to be as compact as possible, while reducing resonances and maximizing the total mass. The feasible choices were tungsten and lead, lead having never before used as an AE sensor backing material. Of the two, lead has a much lower Q, thereby minimizing possible resonance. Lead proved difficult to machine and solder so a lead-based babbitt metal 95% lead, 4% antimony, 1% tin was used, yielding a final backing mass of 20 g. This alloy has a specific acoustic impedance of 25 MPa s/m compared to 34 MPa s/m for the PZT conical element. Brass, a common backing material with specific acoustic impedance very close to that of the piezoelectric element was tried, but exhibited spurious resonances. The backing mass was machined from 9.5-mm round stock, with a 7-mm flat milled along the axis to reduce cylindrical resonances and to provide a mounting point for the FET board. The back end of the mass is cutoff at an angle of 30 to the long axis to reduce symmetry in the design and to prevent direct reflections from reaching the piezoelectric element Proctor, The babbitt metal backing mass is soldered to the piezoelectric crystal using a low melting temperature indium-based solder. Insuring a continuous and uniform bond between crystal and backing is the single most critical aspect of sensor assembly, and a repeatable and dependable construction technique was developed after a long trial and error process. A poor joint is detected during calibration as an anomalous response between 714 and 755 khz. In this frequency range the phase angle shifts by several multiples of 2, accompanied by a strong antiresonance. FIG. 2. Frequency response of the FET-based preamplifier. sor and the A/D converter. The capacitance of 50 coaxial cables are approximately 95 pf/m, while the conical element output impedance is in the tens of pf range Fortunko et al., The amplification system used for the embedded sensor, originally developed by Shiwa et al. 1992, includes a small field emission transistor FET -based Sanyo 2SK715W preamplifier internal to the sensor casing to improve impedance matching. The internal preamplifier also reduces noise pickup in the signal cable, and minimizes the signal losses from capacitive loading in the cable between the conical element and the external amplifier. The circuitry has been optimized to more closely match the characteristics of the smaller conical piezoelectric element Fortunko et al., Figure 2 shows the frequency response of the FET preamplifier. Note that the phase and magnitude are virtually flat from 10 khz to 1 MHz. C. Backing mass D. Casing and acoustic isolation A unique sensor casing was designed so that the sensor would accommodate strain and deformation of the host material, and operate in a brine pore fluid environment under pressure approaching 1 MPa. This has been achieved by pouring a soft Shore-A hardness 9, moldable room temperature vulcanizing RTV rubber between the tubular brass sensor casing and the internals. The RTV rubber provides longitudinal compliance between the sensor casing and internals to maintain intimate contact between sensor tip and specimen i.e., shear spring. The casing constrains the rubber from expanding laterally when a coupling force is applied to the rear of the sensor, allowing the sensor to be placed in a hole just slightly larger than the casing without significant traction forces. RTV rubber has a high Poisson s ratio 0.498, and is relatively insensitive to hydrostatic pressure. Increasing surrounding fluid pressure actually causes the rubber to seal more tightly against water infiltration, whereas pressure proofing a rigid casing would be much more problematic. The surfaces of the sensor internal components are coated with a primer to provide strong adhesion with the rubber, further preventing fluid infiltration. Unfortunately the rubber did not provide acoustic isolation between the host specimen and the sensor internals, resulting in coupling between the backing mass/pzt tip and specimen. III. EXPERIMENTAL METHODS A. Sensor calibration methods Laboratory development of the sensor was guided by frequent calibration. Calibrations were performed on a large 50-mm-thick steel plate, with the signal being measured and the top surface, 76 mm from the source, also input on the plate top Hsu and Breckenridge, Both capillary and 0.3-mm mechanical pencil lead breaks were used as calibration sources per standard AE practice e.g., Breckenridge et al., Although the pencil lead source only approaches a true step impulse, it is convenient to use, very repeatable, and makes for easy comparison with the work of most AE researchers who use this source exclusively. Figure 3 shows a comparison of impulse response of the embeddible sensor to the theoretical infinite plate solution of the same geometry and loading Hsu, The sensor response differs little from the theoretical, with most of the difference due to the lack of 0 12 khz energy that the piezoelectric device cannot sense. The functional difference is the overshoot on the initial spike arrival rebound shown by the sensor. During design and assembly the embeddible sensor re J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1406

4 FIG. 5. Magnitude and phase spectra for the NIST-calibrated embeddible sensor compared with the NIST SRM conical sensor. FIG. 3. Comparison of actual sensor output with results for the same problem solved by ray theory. sponse was also directly compared to the output of a large, brass-backed conical sensor, built and calibrated to NIST Standard Reference Material SRM specifications Glaser and Nelson, 1992a. Final absolute surface calibration of the embeddible sensor was done by the Ultrasonics Standards Group, National Institute of Standards and Technology, Gaithersburg, MD Fick, Figure 4 shows the actual displacement spike measured by the NIST capacitive standard transducer compared with the response of our sensor. Here too, as seen in Fig. 3, there is overshoot on the rebound for the embeddible sensor. The results of NIST calibration of the surfacemounted embeddible sensor is shown in Fig. 5 as frequency magnitude and phase in terms of absolute displacement compared to the frequency response of the NIST SRM conical sensor. Note that the response of our sensor is virtually flat from 12 to 960 khz, with an absolute sensitivity between 28 and 30 V output per nm displacement. For the phase response, the statistical uncertainty of the measurement is greater than the induced phase shift. This calibration shows that the sensor gives an accurate time measure of very small surface displacements over a broad bandwidth. response for the buried sensor surface force Lamb s problem. A sketch of the experimental geometry is shown in Fig. 6. A large 0.33-m cube dolostone block was used as the half-space. The block is homogeneous and isotropic for the scale of the test Nelson and Glaser, 1992, and material properties are given in Table I. The sensor itself was cemented into a 20-mm-diameter borehole 152 mm beneath the horizontal surface upon which the source acted. A hole was bored into the bottom of the block with a flat-end diamond tool, and the bottom handground flat. The sensor is then inserted tip-down into the hole, with the tip resting on the bottom, and the annulus filled with very low viscosity epoxy. Downward pressure on the casing before and during curing results in 1.5-mm axial deformation of the casing relative to the sensor tip. This displacement forces the sensor tip into intimate contact with the rock by the shear-spring action of the RTV rubber in the annular space between the brass casing tube now part of the specimen and the sensing element. The beauty of our design is that the shear spring insures that intimate contact is maintained after cure shrinkage of the epoxy, and specimen deformation. The entire open volume between the sensor and rock is filled so that there is no free surface near the sensing element. The step source used for this experiment was a glass B. Experimental verification of Lamb s problem In order to verify the embedded behavior of the sensor, a set of experiments was run to measure the in situ body wave FIG. 4. Impulse response of the embeddible sensor compared to the NIST reference, NIST Calibration TN FIG. 6. Geometry of Lamb s problem experiment. Note that step force drop was input at epicenter, and every 25 mm up to 152 mm from epicenter J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1407

5 IV. VERIFICATION OF SENSOR PERFORMANCE A. Epicentral results, experiment and theory A rigorous verification process was undertaken to calibrate the embedded sensor to the theoretical particle velocity time histories for the given geometry. The epicentral geometry was used as the fundamental calibration case. The sensor response to a 42-N step force capillary break input normal to the top surface directly above the center-line of the embedded sensor is shown in Fig. 7, with output given as absolute particle velocity time history. For comparison, the theoretical particle velocity time history for this scenario is also shown in Fig. 7. The segments of signals shown covers the period from just before the P-wave arrival to just after the S-wave arrival. Numerous experiments had shown that the dispersive rock effectively filters out frequencies above 450 khz for travel paths much over 50 mm Nelson and Glaser, 1992, so the only processing of the theoretical waveform was 450- khz low-pass filtering single-pole bidirectional Butterworth that mimics the dispersive attenuation of the rock half-space. The experimental time history is presented without signal processing, scaled to theoretical peak first-arrival P-wave velocity. FIG. 7. Comparison of measured and calculated epicentral surface step force of 42 N. The sensor is embedded 152 mm below epicenter and is sensitive to particle velocity normal to the top surface. capillary break rather than pencil lead since the capillary source imparts significantly more force, and better approaches a true step force impulse. For the upper frequency bound of interest to us, 1 MHz, the capillary source yields a near-perfect step drop normal to the horizontal surface Breckenridge et al., NIST typically uses a 0.12-mm o.d. capillary source, yielding a 4-N force step with a s rise time. For our experiment the source pulse must penetrate through more than 215 mm of attenuative and dispersive rock and still have an amplitude well above the noise floor. Therefore 0.5-mm o.d. capillaries were used, yielding a 42-N force, and an estimated rise time of 1.1 s. The first signal was recorded with the step-source in the epicentral position on top of the rock block, 152 mm above the sensor face. The 42-N step force was input on the surface at six successive 25-mm horizontal steps from epicenter Fig. 6. Normalizing linear dimensions by the 152-mm embedment depth, D, the step source was input at 0.167D, 0.333D, 0.5D, 0.667D, 0.833D, and D. All experimental signals were recorded at 14-bit resolution and a digitization rate of 10 mega samples/second. Only the initial portion of each event waveform was used for analysis, from the initial arrival of event energy with the P wave to just after the arrival of the S wave. The use of only the initial, uncontaminated segment of the waveform insured that the analysis would be free from the effects of sample size and shape, and modal analysis would not need to be undertaken. The signals analyzed were effectively from an infinite half-space. B. Results from various geometries, experiment and theory The embedded sensor is designed to be sensitive strictly to motion in-line with the sensor axis. To verify the design, step forces were input on the horizontal top of the test block at various off-epicentral locations hence azimuths relative to the sensor axis Fig. 6. Given a point source point receiver, the radiation pattern for both S- and P-wave motions will have significant energy normal to the plane being monitored. The experimental results were then compared to the theoretical waveforms for the same off-epicentral locations. The comparisons are shown in Fig. 8 a f. The capillary source step force results for offsets of 25 to 152 mm are shown for verification. C. Evolution of the head wave In an elastic half-space there exists for some geometries a refracted shear-wave referred to as a head wave e.g., Pao and Gajewski, 1977; Cerveny and Ravindra, The high fidelity of our sensors was further tested by recording waveforms at source offsets of 107, 112, 117, and 122 mm. Theoretical velocity time histories were also calculated for these geometries. Results of theory and experiment are compared in Fig. 9. Again, the figure shows the waveform from pre-p-wave to post-s-wave arrival, before mechanical resonances of the specimen and source contaminate the waveform. V. DISCUSSION OF RESULTS A. Discussion of sensor epicentral performance and calibration from theory Figure 7 presents the epicentral particle motion time history from the initial arrival of the longitudinal wave P wave to just after the arrival of the shear wave S wave. The rise time of the initial arrival of P-wave energy as measured by the sensor matches the theoretical solution for particle velocity exactly. The behavior at the arrival of the S wave is also in congruence, as is the gradual downward slope between the P- and S-wave arrivals. After extensive theoretical modeling and comparison to the recorded signals, it was decided that the sensor measures particle velocity when embedded, in contrast to its being a 1408 J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1408

6 FIG. 8. Comparison of measured and calculated waveforms due to surface step force. The sensor is embedded 152 mm below the surface. Source located on the surface at a 25 mm, b 50 mm, c 75 mm, d 100 mm, e 126 mm, f 152 mm. particle displacement transducer when surface mounted. When surface mounted, the sensor casing is physically isolated from the half-space. When embedded, it was initially assumed that the soft RTV rubber potting would serve to isolate the active sensor elements from the host motions. However, it was found that silicon rubber has a glass transition frequency at about 10 khz. For example, the Young s modulus of a soft RTV-141 rubber increases almost three orders of magnitude in the 10-kHz frequency range resulting in a high-frequency stiffness of 5 GPa Bosc and Mauguen, We are currently evaluating a urethane elastomer which is quite acoustically isolating at high frequencies Selfridge, The congruence between the measured and calculated particle velocity time histories gives credence to the calibration of the sensor from the Kennett propagator theoretical solution, which accounts for the source kinematics and geometry of the capillary break. Since this is the first time experimental waveforms for the embedded sensor geometry have been recorded, the sensor must be calibrated by comparing experimental sensitivity to the theoretical wave field. Comparison of theoretical particle velocity history and sensor output voltage at the initial P-wave peak yields a particle velocity sensitivity of 2.34 V output per mm/s. The calculation was checked against an independent solution based on a modified Cagniard technique Ruitenbeek et al., 1991, which yields the same calibration factor. The major difference between theory and experiment is the extra downward pulse output by our sensor following the initial peak. Forward modeling could only duplicate this artifact by including a thin, very slow, surface layer underneath the source. Continued improvements in sensor design should eliminate this artifact. The other minor wiggles are assumed due to imperfections of reality, such as finite extent of the source, imperfect material, etc., and do not interfere with interpretation of the recorded signal. The double-pulse 1409 J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1409

7 FIG. 9. Evolution of the head wave, experiment versus theory. The sensor is embedded 152 mm below the surface. Source position on surface a 107 mm; b 112 mm; c 117 mm; d 122 mm from epicenter. should not interfere with quantitative interpretation of source kinematics because it can easily be accounted for through deconvolution. Since the double-pulse appears after the initial rise starts falling, the principle parameters for moment tensor inversion such as polarity of initial P- and S-wave arrival, rise time, and peak amplitude are not distorted. B. Discussion of theoretical and recorded off-epicentral waveforms A comparison of measured and calculated particle velocity time histories for a variety of source receiver geometries is shown in Fig. 8. The sensor was buried in a constant position 152 mm below the horizontal surface, monitoring motions parallel to this exposed surface. A step force was input by capillary breaks every 25 mm from epicenter, from 25 to 152 mm. These six positions correspond to azimuths of 9.5, 18.5, 26.5, 33.7, 40, and 45. As the azimuthal angle increases, more energy is partitioned from the P wave to the S wave, as is evidenced by the changing ratio of P- and S-wave peak amplitudes. For all signals recorded, the rise times and the ratios of P-wave to S-wave amplitude are very similar for theory and experiment. The fit of the S wave is exact for the 75-mm Fig. 8 c offset and within 76% for the 100-, 126-, and 152-mm offsets Fig. 8 d f. For the epicenter and the 25- and 50-mm offsets the Green s function method Johnson, 1974 provided a better fit than the viscous model Kennett, The quality of these results leads to the conclusion that our sensor provides a quantitative time history of the particle motion inside the test block. C. Evolution of the head wave For the geometry used in this experiment, the head wave should appear when the angle between source and receiver was greater than or equal to the critical angle c Pao and Gajewski, 1977; Cerveny and Ravindra, 1971 c sin 1 v s v p J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1410

8 As a test of the accuracy of modeling assumptions made in this paper, a step force was input at four offsets bracketing the expected initial arrival offset. These signals are shown in Fig. 9 a d compared with the actual particle velocity time histories recorded within the test block. For the host material used for this experiment c 37.3, and the head wave should start to appear at a source offset of 116 mm. The headwave first appears at the 117-mm offset compared to the 116-mm offset expected from theory. Note that for both the 117 Fig. 9 c and 122 mm Fig. 9 d offsets the duration and amplitude of the headwave is identical for theory and experiment. VI. APPLICATION OF THE EMBEDDED SENSOR We are working in conjunction with Shell Exploration and Production Research Company, to conduct laboratoryscale experiments of hydraulic fracture propagation Dudley et al., To gain a better understanding of the fracture mechanisms occurring, the test samples are being monitored with an array of acoustic emission and active imaging sensors. Sandstone samples are loaded into a polyaxial load frame located at Shell E&P in Houston, TX, so that independent confining stresses can be applied to the sample in the principle directions, pore water pressure can be controlled, and fluid can be injected into the sample to induce fracture. In this application, the driving consideration for using embedded sensors is the inaccessibility of the specimen surface. During hydrofracture testing at Shell E&P the embedded sensor was subjected to difficult environmental conditions. The test specimen was subjected to confining effective pressures of about 3.5 MPa, and could be saturated with brine with a pore pressure of 1 MPa. Embeddible sensors have been glued inside several specimens which were then loaded to failure and the sensors retrieved by overcoring. The only damage to the retrieved sensors was some tearing of the brass film wearplate. The embedded sensor is an ideal laboratory tool for material science investigations. Potential applications for embedded sensors include: monitoring of concrete dams and highway bridges, monitoring of mines, tunnels, or other geologic structures, and monitoring and basic research on the performance of new polymer and composite structures. Various improvements to the sensor containment are being investigated. The stiffness of the PZT ceramic leads to an impedance matching problem when it is used to monitor relatively low stiffness materials. We are currently investigating other soft elastomers such as urethanes which will not exhibit glass transition at frequencies of interest. VII. CONCLUSIONS Construction details of the embeddible sensor have been presented to give the reader insight into the device. The sensor is based on the NIST conical piezoelectric element, measures particle velocity at a point, and is sensitive only to motion parallel to the sensor long axis. The device is small, 38 mm long and 16 mm in diameter, which allows convenient mounting both inside or on a specimen. This paper presents a unique view of motion inside a solid. For the first time, particle velocity time histories inside a body due to external forces have been accurately recorded. Comparison with theory shows that the sensor is an extremely faithful transducer of particle motion. When embedded the output is proportional to particle velocity and gives a sensitivity of 2.34 V per mm/s. Comparison with theory and NIST calibration prove the sensor to be an excellent transducer of surface displacement with a calibrated sensitivity of 2.8 V per nm. ACKNOWLEDGMENTS Funding was provided by the Gas Research Institute Contract No , Shell Exploration and Production Co., and the National Science Foundation Young Investigator Program Grant No. CMS Many of the manufacturing procedures used to build the embeddible sensor were developed by staff at the National Institute of Standards and Technology, Materials Reliability Division, Boulder, CO. The authors wish to thank all these organizations, and more importantly, the individuals we worked with, for both technical assistance and support. Aizawa, T., Kishi, T., and Mudry, F Acoustic emission wave characterization: a numerical simulation of the experiments on cracked and uncracked specimens, J. Acoust. Emiss. 6, Aki, K., and Richards, P. G Quantitative Seismology Freeman, San Francisco. Bosc, D., and Mauguen, P Acoustic properties of transparent polysiloxanes, J. Appl. Polym. Sci. 40, Breckenridge, F. R Acoustic emission transducer calibration by means of the seismic surface pulse, J. Acoust. Emiss. 1, Breckenridge, F. R., Proctor, T. M., Hsu, N. N., Fick, S. E., and Eitzen, D. G Transient sources for acoustic emission work, in Progress in Acoustic Emission, V, Proc. 10th International Acoustic Emission Symposium, edited by K. Yamaguchi, H. Takakashi, and H. Niitsuma Japanese Society for Non-Destructive Inspection, Tokyo, pp Cerveny, V., and Ravindra, R Theory of Seismic Head Waves Univ. of Toronto, Toronto. Cherry, Jr., J. T The azimuthal and polar radiation patterns obtained from a horizontal stress applied at the surface of an elastic halfspace, Bull. Seismol. Soc. Am. 52, Dudley II, J. W., Shlyapobersky, J., Stanbury, R. B., Chudnovsky, A., Gorelik, M., Wen, Z., Glaser, S., Hand, M., and Weiss, G Laboratory investigation of fracturing processes in hydraulic fracturing, GRI-95/ 0355, annual report to the Gas Research Institute: Chicago. Eisenblatter, J The origin of acoustic emission Mechanisms and models, Acoustical Emission Deutsche Gesellschaft fur Metallkunde e. V., pp Eitzen, D. G., Breckenridge, F. R., Clough, R. B., Fuller, E. R., Hsu, N. N., and Simmons, J. A Fundamental developments for quantitative acoustic emission measurements, Interim Report NP-2089, prepared for Electric Power Research Institute. Fick, S. E Report of Test of Transducers and NIST, Gaithersburg, MD. Fortunko, C. M., Hamstad, M. A., and Fitting, D. W High-fidelity acoustic-emission sensor/preamplifier subsystems: modeling and experiments, IEEE 1992 Ultrasonics Symposium Proceedings, October 1992, Tucson, Arizona, edited by B. R. McAvoy IEEE, New York, Vol. 1, pp Glaser, S. D Development of a high-fidelity, embedded displacement transducer, Proc., 4th World Meet. Acoust. Emission ASNT, Columbus. Glaser, S. D., and Nelson, P. P. 1992a. High-fidelity waveform detection for acoustic emissions from rock, Mater. Eval. 50, J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1411

9 Glaser, S. D., and Nelson, P. P. 1992b. Acoustic emissions produced by discrete fracture in rock-part 2: kinematics of crack growth during controlled Mode I and Mode II loading of rock, Int. J. Rock Mech. 29, Greenspan, M The NBS conical transducer: Analysis, J. Acoust. Soc. Am. 81, Hamstad, M. A., and Fortunko, C. M Development of practical wideband high-fidelity acoustic emission sensors, in Proceedings of SPIE Conference on Nondestructive Evaluation of Aging Infrastructure 1995 SPIE, Philadelphia, Vol. 2456, pp Hsu, N. N Dynamic Green s functions of an infinite plate-a computer program, NB-SIR NIST, Gaithersburg, MD. Hsu, N. N., and Breckenridge, F. R Characterization and calibration of acoustic emission sensors, Mater. Eval. 39, Johnson, L. R Green s function for Lamb s problem, Geophys. J. R. Astron. Soc. 37, Kennett, B. L. N Seismic Wave Propagation in Stratified Media Cambridge U.P., Cambridge, England. Kim, K. Y., and Sachse, W Acoustic emissions from pennyshaped cracks in glass, J. Appl. Phys. 59 8, Kishi, T Acoustic Emission Source Characterization and Its Application to Micro-Cracking, Z. Metallkd. 76 7, Kjartansson, E Constant Q wave propagation and attenuation, J. Geophys. Res. 84, Lamb, H On the propagation of tremors over the surface of an elastic solid, Philos. Trans. R. Soc. London, Ser. A 203, Mal, A. K., Kundu, T., and Xu, P.-C Ground motion calculations, ASME Symposium on Earthquake Ground Motion and its Effect on Structures AMD-53, Michaels, J. E., Michaels, T. E., and Sachse, W Application of deconvolution to acoustic emission signal analysis, Mater. Eval. 39, Miller, G. F., and Pursey, H The field and radiation impedance of mechanical radiators on the free surface of a semi-infinite isotropic solid, Proc. R. Soc. London, Ser. A 223, Nelson, P. P., and Glaser, S. D Acoustic emissions produced by discrete fracture in rock, Int. J. Rock Mech. 29, Ohtsu, M., and Ono, K A generalized theory of acoustic emission and Green s functions in a half space, J. Acoust. Emiss. 3, Pao, Y.-H Theory of acoustic emission, Elastic Waves and Nondestructive Testing of Materials, AMD, edited by Y.-H. Pao ASME, New York Vol. 29, pp Pao, Y.-H., and Gajewski, R. R The generalized ray theory and transient responses of layered elastic solids, Physical Acoustics XIII, edited by W. P. Mason and R. N. Thurston Academic, New York, pp Pekeris, C. L., and Lifson, H Motion of the surface of a uniform elastic half-space produced by a buried pulse, J. Acoust. Soc. Am. 29, Proctor, T. M. 1982a. An improved piezoelectric acoustic emission transducer, J. Acoust. Soc. Am. 71, Proctor, T. M. 1982b. Some details on the NBS conical transducer, J. Acoust. Emiss. 1, Proctor, T. M More-recent improvements on the NBS conical transducer, J. Acoust. Emiss. 5, Rayleigh, Lord On waves propagated along the plane surface of an elastic solid, Proceedings of the London Mathematical Society 17, Ruitenbeek, J. C., Berni, A. J., and de Hoop, M. V Non-far-field vibroseis deconvolution, Publication 1072, Koninklijke/Shell Exp. and Prod. Lab., Rijswijk, the Netherlands. Selfridge, A Acoustic material property tables, Shiwa, M., Inaba, S. H., Carpenter, S. H., and Kishi, T Development of high-sensitivity and low-noise integrated acoustic emission sensor, Mater. Eval. 51, Wadley, H. N. G., and Hill, J. J Dynamic elastic displacements at the surface of an elastic half-space due to defect sources, J. Phys. D 16, Weiss, G. G., and Glaser, S. D Internal measurement of picometer-scale dynamic displacements inside rock, in Transportation Research Record Trans. Research Board, National Research Council NAS/NAE, Washington, DC J. Acoust. Soc. Am., Vol. 104, No. 3, Pt. 1, September 1998 Glaser et al.: Embedded, wideband velocity sensor 1412