In recent years, there is an increasing interest in structural condition monitoring technologies capable of detecting structural damage at the early stages. Early damage detection can be viewed from Non-destructive Evaluation (NDE) and Structural Health Monitoring (SHM) perspectives that, although retain many common elements, reflect different philosophies, technical realizations and damage assessment practices. A broad spectrum of NDE methodologies is oriented towards shortterm periodic inspections realized with repositionable transducers, signal condition units and other hardware modules. Although a number of promising early damage detection NDE techniques have emerged in the past decade, thermography [1] and positron annihilation [2] among others, the following discussion is primarily focused on exploring recent advances in structural health monitoring of the incipient damage.

The philosophy of the structural health monitoring is strongly influenced by the bio-mimetic analogies with selfsensing biological systems featuring a synergistic integration of sensors, signal transmission lines, analysis By elements and information inference/classification units. Structural health monitoring systems adapt a similar architecture in which sensing elements are configured in sensor networks with signal processing, decision making and status reporting capabilities.

One of the critical aspects in operating complex structural systems is ability to assess their current and projected condition. Unnoticed structural deterioration may potentially compromise structural safety and lead to a catastrophic event resulting in personnel risk and/or financial losses. To minimize risks and losses, SHM systems are integrated into key structural elements for continuous structural monitoring. SHM provides near real-time indication of structural health that is used by designated personnel to make a final diagnostic decision typically falling into three categories: structure is healthy, structure needs attention, and structure is damaged. A diagnostic decision represents a final product of human-machine information exchange and analysis. Although final result of structural assessment may be represented on a simple and universal red-yellow-green scale, analysis methods depend on a damage scenario and could incorporate sophisticated measurement and information inference practices. A diagram on Figure 1. illustrates major elements of the diagnostic process. Sensors are selected for investigation of a particular damage phenomenon. Depending on damage physics and its manifestation via selected sensing mechanism, a measurement method and associated instrumentation are chosen. The output of measurement equipment often is digitized and then is analyzed using a spectrum of signal processing techniques to obtain data features indicative of damage. The decision support algorithm, which could be implemented with neural networks, statistical analysis and etc., classifies signal features into categories associated with structural condition, i.e. “healthy”, “damaged.” In the last phase of structural condition assessment, the SHM system output is evaluated by the operator, who makes a final diagnostic decision.

Figure 1. Major elements of structural diagnostic process

An ultimate goal of structural health monitoring and condition assessment is to detect, locate, and if possible classify an incipient damage. Selection of an appropriate SHM methodology is ultimately related to type, nature, and severity/extent of potential damage. Depending on a scale of the consideration, the damage may be viewed as a continuously increasing level of disintegration progressing from material to structural scales. A diagram in Figure 2 presents examples of incipient damage corresponding to different physical scales. An attribute “incipient” is relevant to a global aspect of the structural health. In other words, monitoring of the fatigue-induced dislocations dynamics and formation of micro-cracks in a turbine blade implies quite a different definition of the “incipient” damage in comparison with the integrity assessment of the satellite's bolted joints.

The damage could reveal itself through a broad spectrum of physical phenomena. Figure 3 presents a pictorial representation of this generalization and, in effect, suggests possible energy transformations aiding the detection. Examples include variation of electrical conductivity visualized using embedded nano sensors [3] or evolution of thermal properties monitored with infrared camera [1]. Our primary interest, however, lies in identifying and characterizing changes in mechanical parameters caused by damage. We consider dynamic mechanical methods that encompass condition monitoring approaches ranging from low frequency structural vibrations to an ultrasonic microscopy. Analysis of the frequency content of the elastic disturbance utilized in the mechanical methods allows for their convenient presentation in terms of several clusters (Figure 4). Naturally, a distinction could be made between active and passive technologies. Active techniques interrogate the test structure or the structural element with an elastic signal of a particular form and information on the presence of damage is inferred from the analysis of the structural/material response. Passive technologies listen to indications of the damage-initiated mechanical events such as acoustic emission or changes in stress levels. Although formulation of “incipient” is rather broad, it bears the “local” context. It is assumed that local incipient changes of the mechanical parameters may not introduce substantial variation of the global dynamic signatures, which increases complexity of the damage detection task.

Figure 2. Examples of damage at various scales

Figure 3. Physical manifestation of material damage

Figure 4. Dynamic mechanical NDE/SHM methods

Acoustic Emission

Acoustic emission (AE) has long been used for continuous monitoring of structural deterioration and integrity of structural connectors and joints. AE is a passive technique that listens to the acoustic activity triggered by external loads exerted on a structure or a material sample. The origin of the AE signals, or “acoustic noise”, significantly depends on the nature of the studied material. In metallic samples, AE sources may include dislocations' dynamics, coalescence of micro-cracks, formation and growth of macro-cracks and/or chemically triggered events (e.g. corrosion). In the composite material, the AE events may be related to disintegration of the matrix, fibers and delaminations along the matrix-fiber inter face. Depending on geometrical configuration of the structural element, the AE event produces an array of wave modes that is typically detected with a broad-band AE transducer covering 0.1-1 MHz frequency range. In AEbased structural health monitoring applications, correlation between a number of AE occurrences and incipient material damage is sought to establish a damage detection threshold and define classification margins for the artificial intelligence (AI) information inference and the decision support.

Early investigations on the use of AE signals for deducing information on structural condition can be traced to late 50s [3]. However, it is not until the revolution in acoustic instrumentation brought by affordable digital data recorders and automatic signal analysis algorithms that the AE technique received considerable attention of engineers and researchers working in the nondestructive evaluation field. Monitoring of the AE activity often implies answering two questions: how many AE events (signals) occurred and what are the parameters of these signals. These questions essentially define requirements for the data analysis procedures addressing both detection and classification of the AE event.

According to Surgeon and Wevers (1999) [4], three types of data analyses are applied to AE signals: AE activity analysis, AE parameter analysis, and AE frequency analysis. The activity analysis implies setting up a detection threshold and monitoring evolution of damage. As a result, structural deterioration charts can be constructed that relate a number AE events and monitoring parameters such as stress magnitude or number of fatigue cycles. Moving the AE activity analysis one step forward, a variety of AE signal parameters (number of counts, signal energy and duration, peak amplitude, etc.) have been suggested to facilitate inference of information pertaining damage type and severity. For example, [5], [6] has showed that analysis of signal parameters is instrumental in identifying different stages of crack development. Spectral characteristics of the AE signal give a different perspective on the AE event identification. The frequency analysis allows for determining wave propagation modes and thus providing additional information on the AE source. Initial studies employed the power spectrum representation. In recent years, however, a timefrequency analysis was recognized as an effective tool for studying spectrum's evolution. For this purpose, both the spectrogram and wavelet analysis have been extensively used [7].

Structural health monitoring systems utilizing AE principles have been successfully employed in aerospace [8], pressure vessel [9], offshore [9] and many other applications. However, limitations of the AE SHM preclude an extensive use of this damage monitoring approach. The limitations include contamination of the AE signal by mechanical, thermal, and other noise sources, difficulties associated with source location (attenuation, material anisotropy, sensor location and coverage, etc.), complexity of the signal interpretation and AE event identification. Miniaturizing the AE sensors and instrumentation in the context of addressing the low signal amplitude problem also need to be addressed. It needs to be mentioned that current AE research is directed towards resolving or mitigating effects associated with the indicated limitations.

Ultrasonic testing is probably the most popular nondestructive evaluation and condition assessment methodology. A broad spectrum of ultrasonic techniques was developed over more than half a century to facilitate nondestructive inspection of metallic and non-metallic materials and structures of complex and simple geometries. Most of the existing techniques may be separated into “through transmission” and “echo” operation modes. In the first approach, information about the damage is inferred from the parameters of the signal propagated through the damaged zone and picked up by the second transducer. The second approach utilizes only one transducer that is switched between the transmission and reception modes. The transmitted elastic wave is reflected by the structural damage and presence of the reflected signal indicates extend and severity of damage. Conventional procedures for ultrasonic NDE utilize contact transducers that are installed on a structural element during inspection. In contrast, continuous SHM is realized with small unobtrusive embeddable sensors that are permanently bonded to the host structure. Crawley and Luis [10] were among the first to suggest that thin piezoelectric wafers can be used for structural excitation and control. Since this early work, many researchers have been using embeddable piezoelectric patches, or piezoelectric wafer active sensors (PWAS), for a variety of applications including SHM. The elastic wave propagation methodologies based upon transmission and reception of ultrasonic waves by PWAS received considerable attention under the umbrella of “embeddable ultrasonics” [11-19]. While it is impossible to discuss all outstanding contributions to this rather broad field, we would focus on studies that show capabilities of the embedded ultrasonics in detecting and characterizing an incipient damage.

Embeddable ultrasonics is a valuable alternative to traditional wave propagation methods as it allows for very fast long range inspections, monitoring of inaccessible areas in structural assemblies, and continuous assessment of the structural integrity. Practical applications of embedded ultrasonics range from aerospace to mechanical and naval structures. Specifics of the structural geometry often implies utilization of the guided wave propagation modes rather than the bulk wave inspection. In the guided waves propagation, dispersion characteristics need to be accounted for while selecting a particular wave modes. To reduce complexity of the condition assessment system, the non-dispersive modes are preferable. In practice, the S_o symmetric mode exhibiting low dispersion at low frequencies is selected as oppose to the dispersive flexural wave A_o. Figure 5 illustrates waveforms of both modes in a thin aluminum plate. Although exploring relatively low frequency range has advantages of the non-dispersive long-range propagation, sensitivity of the ultrasonic wave to the small-scale defects is not exceptional at these frequencies. The best detection is achieved when the wavelength of the transmitted wave exceeds size of the monitored in homogeneity. For frequencies below half a MHz this consideration results in the minimal detectable defect size of the order of 5 mm. However, continuous comparison of the measured data with a baseline allows to push this limit even further. For example, Yu and Giurgiutiu (2005) [20] reported detection of a pin hole as small as 1.57 mm. To achieve high sensitivity to the smallscale defects, a consecutive series of signal conditioning procedures including hardware filtering, multiple averaging, time-frequency analysis, and envelop detection is suggested.

Figure 5. Embedded ultrasonic sensor network configured on a thin aluminum panel and a typical wave propagation signal

Real-time monitoring of the fatigue crack growth with embedded ultrasonics were discussed by Ihn and Chang (2004) [13] and Giurgiutiu et al., (2006) [Error! Reference source not found]. Ihn and Chang (2004) applied the through transmission (pitch-catch) method for the detection of fatigue cracks in 462 mm × 936 mm aluminum single lap joint panel. The sensor strip was mounted next to rivets and 420 kHz signal was analyzed at consecutive intervals of the fatigue load spanning from 40 kcycles to 160 kcycles. The system reliably detected cracks longer than 7 mm.

Identification of fatigue damage using the embedded ultrasonics structural radar (EUSR) was reported in [Error! Reference source not found]. A 372 kHz S_o Lamb mode was used as an interrogation signal. A pre-crack of 30 mm long and 125 µm wide was made in a 700×600×1 mm 2024-T3 aluminum plate to initiate crack development during fatigue. The results of the test have demonstrated ability of the EUSR embedded ultrasonics to detect damage progressing from 30 mm to 60 mm. However, authors suggested that for crack lengths exciding the EUSR size, the aperture effect of the ultrasonic radar needs to be taken into account.

Studies on monitoring damage initiation and progress using the embedded ultrasonics have shown potential of this SHM approach for detecting incipient damage of the order of several millimeters. This is also consistent with investigations for composite materials [22,23] suggesting utility in identification of delaminations of approximately 10 mm long. The detection threshold of the embedded ultrasonics may be improved by selecting higher excitation frequencies, which unfortunately leads to pronounced manifestation of dispersion effects.

Discussion in the previous section suggests that considerable progress has been achieved in both fundamental and applied research on utilizing the elastic wave propagation for SHM. It has been shown that this approach allows for detection and location of structural damage ranging from delaminations and disbonds to small cracks. However, detectability of structural damage in the elastic wave propagation techniques significantly depends on the wavelength of the transmitted/received elastic wave [24]. On one hand, very small defect may not be resolved adequately if relatively low frequencies are employed. On the other hand, high frequency ultrasonic wave may attenuate rapidly imposing limitations on the effective detection range. In complex structural components misinterpretation of the test results is possible if the damage and structural feature (a hole, notch, etc.) reflects the same amount of the acoustic energy.

One possible way of addressing the small-scale damage and misinterpretation issues is to explore the nonlinear nature of the structural damage. It is worth mentioning that, correlation between the acoustic nonlinearity and material damage is known for over forty years [25]. This work was pioneered by Mason [26] in direct connection with studying physical properties of solids. Generation of ultrasonic second and third harmonics due to the microscale damage (dislocations) was discussed by Hikata, Elbaum, and Sewel in their classical papers presented in mid sixties [27-28]. Since then, many authors contributed to this subject. While it is impossible to provide a complete list of contributions, examples include work of Breazile [29], Cantrell and Yost [31], Nagy [31], Van Den Abeele et al. [32], Rokhlin [33]. Recent results on the use of the nonlinear acoustic approach for assessment of degradation in the aerospace materials were reported by Frouin et al. [34], Na et al. [35], Doskoy et al. [36], Zagrai et al. [37]. The LANL group headed by Johnson uncovered an anomalously high mesoscopic nonlinearity [38] that in the material with multithestructural damage may orders of magnitude exceed the ”classic” elastic nonlinearity. Although, there is a considerable volume of work on application of the principles of the nonlinear wave propagation in NDE, realization of this methodology in the SHM context has been limited [39]. Only recently, the nonlinear methods became an active discussion topic in the SHM community [40]. The development in this field is currently dominated by the nonlinear SHM methodologies exploring structural dynamics at relatively low frequencies [41-43].

Depending on physical realization, high frequency nonlinear ultrasonic methods may be divided into three distinct groups: the resonance method, the finite amplitude method and the modulation method (Figure 6).

Figure 6.Nonlinear acoustic damage assessment (a) resonance method, (b) finite amplitude method, (c) modulation method.

In the resonance method, the structural condition is assessed based on a shift of a resonance curve at increasing levels of the applied excitation. This methodology was introduced by the LANL group headed by Johnson [38] for studying nonlinear effects in geomaterials. Many other researchers contributed to further development of the method including [32] and others. The resonance method explores global structural response at relatively low frequencies and may be realized in structures with pronounced resonance properties. This, however, imposes certain limitations on range of potential applications.

When a harmonic signal is introduced in a nonlinear medium such as material with cracks, the frequency analysis of the response reveals additional spectral harmonics. Correlation between the amplitudes of these harmonics and extent/severity of the material/structural damage is well known and is used in the finite amplitude method [27,30,34]. The finite amplitude method presents a convenient avenue for studying the nonlinear phenomena as direct analytical formulations are available that connect the nonlinear parameter in the second order nonlinear wave equation and amplitudes of second and other harmonics in the spectrum of the measured acoustic response. Among other fundamental studies, the finite amplitude approach was instrumental in experimental investigation of the nonlinear acoustic manifestation of the incipient fatigue damage [30,34,35]. While Cantrell and Yost [30] obtained the nonlinear acoustic parameter for the fatigued aluminum alloy in the off-line mode, Frouin et al., [34] monitored change in the amplitude of the second harmonic (and thus a nonlinear parameter) during a fatigue test. A continuous increase of the nonlinear acoustic parameter due to fatigue was observed with a flat plateau formed before the onset of the macro-scale crack nucleation suggesting that the nonlinear ultrasonic measurements may be potentially used for continuous assessment of structural health before fracture. However, at that point, implementation of this methodology in a SHM regime was not considered as authors focused on fundamental understanding of behavior of the nonlinear acoustic parameter during the real-time fatigue experiment. It needs to be mentioned that potential disadvantages of the finite amplitude method include rapid attenuation of the propagating higher harmonics and detrimental contribution of the equipment nonlinearity into the same spectral components utilized in assessing the structural damage.

Contribution of the equipment nonlinearity into the measurement results may be reduced by utilizing a multi-frequency interrogation signal. Such a signal realization forms the basis of he modulation measurements that provide an alternative way of measuring the material nonlinearity. It is likely that original development of the modulation technique was carried out by a group of scientists at Institute of Physics, Nizhniy Novgorod, Russia [44]. The technique rapidly became popular for applications ranging from detection of cracks [33, 45,46], assessment of delamination-debonds, monitoring fatigue damage, and evaluation of structural joints. The multi-frequency excitation signal utilized in the modulation technique contains both low and high frequency components. The low frequency vibrations expand and contract damage interfaces effectively altering the propagation conditions for the high frequency ultrasonic wave. This leads to modulation of an ultrasonic wave by low frequency vibrations at the site of damage. The resultant side-band (modulation) components in the spectrum of the acquired ultrasonic signal indicate presence of damage. Work of Donskoy et al., (2001) [45] confirmed direct correlation between amplitudes of the side-band spectral components and damage severity. The modulation index, defined as a ratio of amplitudes of the sideband and fundamental frequency components, was measured for steel beam specimens containing fatigue cracks. Gradual increase of the modulation index was observed for specimens with increasing crack depths.

Application of the modulation technique for assessment of adhesive bonds was reported by Rokhlin et al., [33].In the degraded bond, nonlinear interaction of the ultrasonic pulse with the low frequency vibrations resulted in a dynamic shift of the resonance frequency of the bond layer. The later was used as a damage detection feature and correlated well with the bond quality. Yan and Nagy [45], developed a modulation method based on the nonlinear interaction of ultrasonic interrogation waves and laser-induced thermal stress pulses. Authors utilized thermal stresses for achieving the crack's “breathing” condition to facilitate modulation of the ultrasonic pulses. It has been shown that the proposed approach could (a) detect a surface crack and (b) discriminate crack from surface scratches, corrosion pits, etc. One of the disadvantages of the laser-based modulation technique is that high intensity laser pulses were needed to achieve thermal stresses that could induce crack breathing.

Development of the vibro-modulation SHM approach was reported in [39]. Embeddable piezoelectric wafer active sensors were utilized to detect and discriminate damage from structural features such as holes. Low frequency vibration was delivered by a magneto-strictive shaker attached to an aluminum beam. Piezoelectric sensors were excited with CW signal covering 224-244 kHz frequency range. Results of the experiment are presented in terms of linear and nonlinear responses in . The linear response was measured at the fundamental frequency and the nonlinear response was obtained at the sum frequency (ultrasonic + vibration). A typical Fourier spectrum for a particular excitation frequency is also depicted in the figure. Noticeably, the modulation level obtained from the specimen with a fatigue crack is much higher than the modulation measured in the specimen with no damage and a specimen with a hole. Average modulation indexes calculated from the figure (crack -50.66 dB, hole -67.43 dB, intact -66.46 dB) indicate damage selectivity of the modulation method. Experiments confirmed applicability of the vibromodulation method for the SHM involving permanently bonded unobtrusive piezoelectric active sensors.

Figure 7. Linear (fundamental frequency) and nonlinear (sum frequency) responses and typical spectra measured for (a) aluminum sample containing a fatigue crack, (b) intact aluminum sample without crack, (c) sample with a hole. (Adapted from [38])

Application of the modulation approach for assessment of fatigue damage before an onset of macro-scale cracking and structural disintegration was presented in [47]. Piezoelectric transducer were installed on an aluminum beam subjected to a three-point-bending fatigue test. The high frequency ultrasonic sweep signal was realized for the selected transducers. Low frequency vibrations at 10 Hz were delivered via fatigue machine. The experimental setup allowed for obtaining a measurement of the nonlinear acoustic modulation index every 180 fatigue cycles. Results of the real-time vibro-modulation assessment are depicted in Figure 8. The damage index was calculated on a linear scale as a difference between modulation indexes correspondent to intact and damage conditions of the same sample. A stable increase of the nonlinear acoustic damage index was observed up to a crack initiation which introduced instability in the data record. Simultaneously, mechanical parameters of the test sample were monitored via dataflow from the testing machine and, as a result, a relative compliance was calculated and presented versus the measured nonlinear damage index. Therefore, a correspondence between decrease of the specimen's stiffness due to fatigue and the nonlinear acoustic damage index was obtained experimentally. Further studies [36,37] revealed direct correlation between values of the damage index and the micro-scale material damage imaged using acoustic microscopy and SEM. Therefore, it is advocated that the nonlinear acoustic modulation approach may become a valuable method for SHM of incipient damage before and after material fracture.

Figure 8(a) Nonlinear acoustic damage index versus number of fatigue cycles and (b) relative compliance versus damage index for the aluminum specimen fatigued to failure. (Adapted from [46])

The Electro-Mechanical Impedance (EMI) method takes advantage of a localized dynamic interaction between the host structure and a piezoelectric sensor that facilitates manifestation of the structural dynamic features in the electrical response measure at the sensor terminals [48,49,50]. The mechanism of the EMI method can be explained as follows [51].

Figure 9. Electro-mechanical coupling between the active sensor and the host structure

Under electrical excitation, the bonded piezoelectric active sensor produces local strain parallel to the structural surface as indicated in Figure 9. The reaction of the host structure to this excitation can be presented in terms of the structural impedance with longitudinal and flexural components incorporating equivalent mass, stiffness, and dissipation elements [51].

[1]

Due to the mechanical interaction between the sensor and the host structure, the structural impedance affects the sensor impedance

[2]

and, through the electromechanical coupling inside the active element, is reflected in the electrical impedance measured at the sensor's terminals.

[3]

In the expression (3), describes the mechanical characteristics of the system transformed into the electrical response using the transformation factor N² . Figure 10 shows the details of the equivalent electrical circuit describing dynamics of the sensor and a host structure. Formulation (3), allows for comparison of the analytical model and the experimental data. An example of such a comparison is presented in Figure 10, in which peaks of the experimentally obtained admittance match with theoretically calculated structural resonance of an aluminum beam. Noticeable in the figure, a slight discrepancy between theoretical and experimental results for flexural modes at high frequencies may be attributed to limitations of the considered Euler-Bernoulli beam model.

Figure 10 (a) An equivalent circuit diagram describing the sensorstructure interaction in the electro-mechanical impedance structural identification; (b) experimental and calculated electromechanical admittance spectra of the free-free aluminum beam

Availability of structural dynamic signatures through the impedance measurements allow for realizing the continuous monitoring of structural condition. An incipient damage affects local mechanical characteristics that lead to change of the impedance magnitude, redistribution of resonances and appearance of additional peaks. These variations form a basis for feature vectors used in the artificial intelligence algorithms for damage identification and classification as illustrated in Figure 11. Sensitivity of the electro-mechanical method depends on the excitation frequencies of the piezoelectric sensor. It has been shown [50] that the impedance method perform effectively in the frequency range spanning from just a few kHz to hundreds of kHz.

Figure 11. Electro-mechanical impedance structural health monitoring employs a network of piezoelectric active sensors, impedance measurements, response comparison and classification

The electro-mechanical impedance method has been used for structural health monitoring of the incipient damage since early 90s. Chaudhry et al., (1994) [52] described the use of the E/M impedance technique for health monitoring of a three-bay aluminum truss representing a space structure. Information on structural condition was inferred from comparison of the impedance spectra of damaged and undamaged scenarios. In the experiment, the damaged condition was simulated by loosening one of the member's connection with the nodal derlin-ball. Potential use of the EMI method for local-area health monitoring of a tail-fuselage aircraft junction was explored by Chaudhry et al., (1995) [53]. Large damage readings were recorded for the smallest bolt turn in the near field, while almost no reading was obtained when the same change was applied to a bolt in the far field. It was suggested that the method is highly sensitive to actual damage, while it is relatively insensitive to other types of changes taking place during the normal operation of the aircraft. Studies on utilizing the EMI method for monitoring structural connectors and joints have been reported by Giurgiutiu et al., (1998), Park et al., (2001, 2005), Pears et al., (2004) [53-57] and have shown effectiveness of the method for monitoring an instant loss of structural integrity and feasibility of automatic condition assessment. Childs et al., (1994) [58] have demonstrated that the electromechanical impedance signatures are useful in monitoring condition of complex precision parts. The method was able to recognize cracks located near and further away from the sensor bonded to a high precision gear. The crack adjacent to the sensor produced considerably larger value of the impedancebased damage index than the distant crack. Sensitivity of the EMI method to the simulated cracks in circular plate specimens and realistic aircraft panel has been studied in [49]. Experiments revealed that although the crack was detectable more than 5 cm away from the sensor, significant changes in the impedance spectrum occurred for the crack located only 1-2 cm from the sensor. In the aircraft skin specimens, the crack located several centimeters away from the sensor was detected using a dense impedance spectrum and probabilistic neural network processing. Applications of the electromechanical impedance method for monitoring civil infrastructure were reported by Quattrone et al., (1998), Park et al., (1999) and most recently by Bhalla and Soh (2004) [59-61] [59-61].

The Magneto-Mechanical Impedance (MMI) technique bridges together structural dynamic testing and magnetic field measurements yet differs from the Electro- Mechanical Impedance [50] and the Eddy currents inspection methods [24]. In MMI, a magneto-elastic generation of elastic waves [62] via a coil and a magnet (Figure 12a) is utilized to excite structural vibration modes simultaneously sensed by the same Magneto-Elastic Active Sensor (MEAS).

The structural excitation and sensing mechanism is based on the phenomenon described by Banik and Overhauser (1977), [63]. The elastic waves generated by magnetoelastic sensor (Figure 12a) travel in the medium and reflect off the boundaries producing standing wave (modal) spatial patterns at the respective natural frequencies. Sweeping the transducer's excitation frequency allows for obtaining the dynamic impedance of the sample, which uniquely represents its dynamic behavior (structural resonances/antiresonances) in the measured frequency range. It can be shown [64] that the impedance measured with MEAS is described by the following expression.

[4]

Where ω is the excitation frequency, L_MEAS and R_MEAS are inductance and resistance of the sensor, L_S is inductance of the structure and k_C is the electro-magnetic coupling coefficient 0 ≤ k_C ≤ 1. Structural impedance Z_str (ω) in equation (4) accounts for dynamic characteristics of structure and hence may reflect structural changes due to damage. By examining changes in the MMI spectra one may assess structural condition and infer information on potential damage.

Parameters in equation (4) are derived from the circuit presented in Figure 12b. This equivalent circuit indicates contribution of structure Z_str (ω) and the sensor (MEAS) in the total impedance. Electro-magnetic coupling between MEAS and structure is accounted for using the mutual inductance M. Representing structural excitation as a Lorentz force acting at the point x_a of an elastic beam , F_L (x,t)=lBb(x-xa) e^iwt one arrives to the following , expression for structural impedance:

[5]

where ω is excitation frequency, b_a is MEAS's lateral dimension, ω_n represent natural frequencies, W_n (x) are mode shapes, and are damping ratios. For a freely supported thin aluminum beam of length 304.8 mm, width - 25.4 mm, thickness - 1.587 mm, modulus of elasticity 73.1 GPa, and density 2780 kg/m³ , Eq. (5) combined with Eq.(4) incorporating electrical parameters, L_MEAS ≤ 1.9 mH and R_MEAS ≤ 10 Ω, L_s = 0.1µH and k_C ≤ 0.4, yield a theoretical curve ( blue) depicted in Figure 13. It should be noted that the analytically calculated impedance shows a slope proportional to iωL_MEAS modulated by impedance peaks associated with flexural vibrations of the aluminum beam. The model was validated in MMI measurements of the aluminum beam using HP 4192A impedance analyzer and the MEAS sensor. An experimental curve (red) is also illustrated in Figure 13. As it could be seen from the figure, theoretical and experimental responses match reasonably well. The inductive slopes are comparable and position of impedance peaks matches closely. Discrepancies in the latter may be attributed to limitations of the Euler-Bernoulli beam theory at high frequencies. Figure 13 indicates that structural response associated with impedance peaks may be inferred from MMI measurements. Consequently, in the MMI nondestructive assessment method, the damage affects the structural impedance and the impedance changes are reflected in structural MMI response.

Figure 12 (a) Schematics of the magneto-elastic active sensor; (b) equivalent circuit for magneto-elastic active sensing

Figure 13. Theoretically calculated and experimentally measured MMI responses of an aluminum beam

Preliminary results showing potential of the proposed approach are presented in Figure 14. Experiments were conducted with an improvised first generation magnetoelastic probe that consisted of a coil and a magnet. A HP 4192A impedance analyzer was used to acquire the data via GPIB inter face coupled with Labview. The experimental specimen used in a proof-of-concept test was a painted aluminum honeycomb panel with an introduced edge disbond. Figure 14 shows the MMI spectra obtained for the healthy (labeled A and C) and damaged (B) areas of the panel. The Z_trend =R+iωL dependence was removed from the data for clarity of presentation. Comparison of the spectra reveals redistribution of impedance peaks and appearance of new anti-resonances in the spectrum of the disbonded area. It is suggested that such a variation of the MMI spectra may be used for damage detection and classification.

Figure 14. Magneto-mechanical impedance detection of a disbond: (a) experimental set up of the proof-of-concept MMI test, (b) results of the experiment showing impedance signatures of healthy and damaged portions of a honeycomb panel

Structural deterioration under environmental and operational loads may be assessed using a broad spectrum of SHM methods. In particular, we discussed local damage detection approaches based on excitation and/or reception of elastic waves. Damage specifics, ambient noise, detection range, and structural complexity are among factors governing selection of a suitable SHM technology. Definition of the incipient damage depends on an application paradigm. One example is a micro-sale damage in the form of dislocations conglomeration and micro-cracking developed before initiation the macro-scale fracture. Such an incipient damage may be assessed nondestructively via AE testing or nonlinear acoustic methods. However, these methods manifest susceptibility to ambient noise and all contributing effects leading to the deteriorated performance need to be eliminates before testing. The embedded ultrasonics allows for inspection of relatively large structural areas, but the minimum detectable damage size is in the order of a few millimeters. Impedance techniques manifest low sensitivity to defects in the far field, but enable ident i f icat ion of damage in the immediate neighborhood of the sensor. They are also sensitive to delaminations/disbonds and some realizations (MMI) may allow even for the non-contact operation.

It is worth mentioning that one of the promising trends in SHM is utilization of alternative sensor modalities without significant increase of additional system components. In other words, one sensor is employed as an active (or passive) element in several SHM approaches. This synergistic integration of diverse damage detection algorithms allows to form “orthogonal” identification features and thus improves damage detection and characterization capabilities. For example, piezoelectric wafer active sensors may support embedded ultrasonics, electro-mechanical impedance, and passive impact detection schemes. From the technological point of view, advances in miniaturization of system's components are desirable as they facilitate considerable weight and system integration savings. Another technological avenue encompasses improvements in material performance and reliability. These innovations enable a next step in exploration of new physical effects for damage detection and characterization.

Author would like to acknowledge NMIMT administration and US Air Force Office of Scientific Research (AFOSR) for sponsoring a portion of this study.