Insight Vol 52 No 2 February 2010 61

ROY SHARPE PRIZE LECTURE 2008

Electromagnetic, thermoelectric and other non-destructive evaluation methods offer unique opportunities for materials’ state awareness monitoring. A variety of sensors can be built based on these principles to detect and quantitatively characterise subtle environmentally-assisted and/or service-related changes in the state of metals, such as microstructural evolution, phase transformation, plastic deformation, hardening, residual stress relaxation, increasing dislocation density, etc. In most cases, the detection sensitivity is sufficiently high for the purposes of materials’ state awareness monitoring and the feasibility of the sensing method is mainly determined by its selectivity, or the lack of it, to a particular type of damage mechanism. This paper reviews recent developments in electromagnetic and thermoelectric materials characterisation and offers a feasibility assessment of these non-destructive methods through a couple of representative examples of broad interest.

Keywords: ACPD, eddy currents, electrical properties, condition monitoring.

1. IntroductionThe future paradigm of condition-based maintenance (CBM) has been the Holy Grail of non-destructive evaluation (NDE) and structural health monitoring (SHM) for many years. Significant reduction in the ability of a component to perform its function and its subsequent failure generally begin at the material level, therefore successful implementation of CBM hinges on full characterisation of the current state of material, which is referred to as materials’ state awareness (MSA)[1]. Figure 1 shows a schematic representation of the challenging road from CBM to MSA through the intermediate steps of future state prediction and failure prediction. Key elements of this strategy are reliable and accurate predictions of the future state and expected failure modes of the material system based on its current state and expected usage. The present paper will attempt to summarise the main opportunities and challenges facing the NDE community in the crucial first step of MSA, which incorporates such diverse principles as structural analysis, materials science, applied physics and sensor development.

The main problem with such an ambitious plan, of course, is that one must know literally everything about the current state of the material to have any chance of success in the follow-up steps.

For example, the following quote from the Discovery Challenge Thrust issued by the Air Force Office of Scientific Research (USA) summarises the ‘requirements’ for achieving MSA in aerospace materials, components, structures and systems: “Determine in real time the current state so that the remaining capabilities of the system or component can be predicted with a high degree of accuracy and known level of confidence (i) for any material systems and material processing, (ii) for any operational environments, component usage history, and (iii) for any failure or material/structure/system degradation mode”[2]. These general requirements are not application-specific and could be adapted to other fields, such as the nuclear industry, where state-of-the-art materials are used close to the very limits of their capabilities and reliability during extended service in hostile environments is of the utmost importance.

The technology challenges associated with this multidisciplinary research effort are (i) to assess early and progressive changes in material state associated with operational usage and exposure, (ii) to predict the real-time physical, chemical or electronic state at specific locations for complex systems subject to service loads and environmental exposure over time, and (iii) to relate the current and evolving state of microstructure and damage processes to enable probabilistic prognosis modelling of the material/structural/system state[2].

2. Materials’ state awareness monitoringInspection methods for MSA will have to be based on simple, reliable and rugged non-destructive monitoring procedures that can be easily adapted to field conditions. This ultimate goal can be best achieved by screening the most adaptable NDE techniques for sensitivity and selectivity rather than starting with the most sensitive and selective known laboratory techniques and later trying to adapt them to in-situ monitoring. Ideally, MSA sensors

Non-destructive methods for materials’ state awareness monitoring

P B Nagy

Peter B Nagy is with the Department of Aerospace Engineering, University of Cincinnati, Cincinnati, Ohio 45221, USA and the UK Research Centre in Nondestructive Evaluation, Imperial College, London SW7 2AZ, UK. Email: peter.nagy@uc.edu / p.nagy@imperial.ac.uk

The Roy Sharpe Prize is for significant contribution through research and development in any branch of NDT to the benefit of industry or society. This prize is awarded by the Technical Committee of the Institute and sponsored annually by an organisation in Associate membership of the Institute. The 2008 Roy Sharpe Prize was funded by TWI Ltd.

Based on the Roy Sharpe Prize Lecture presented at NDT 2009, the 48th Annual British Conference on NDT, Blackpool, UK, September 2009

Figure 1. A schematic representation of the road from condition-based maintenance to materials’ state awareness

DOI: 10.1784/insi.2010.52.2.61

should have a low profile, be of robust construction, be unaffected by the operating environment and have simple wire connections. Electromagnetic methods are well suited for this purpose. Here, electromagnetic methods will be defined in the most general sense as those measuring electric signals produced by electric, magnetic or thermal stimulus. The particular methods to be considered in this paper all have established track records in non-destructive assessment of materials degradation in a wide range of engineering materials and are particularly well suited for monitoring thermo-mechanical damage processes such as fatigue, creep, embrittlement, hardening, etc. Figure 2 illustrates the complex relationships between material and component state parameters and measurable electromagnetic physical properties. The measured properties are called either intensive or extensive depending on whether they are independent or dependent on geometrical properties such as size and shape. Even with the above-introduced very general classification, there are only four fundamental intensive physical properties we can exploit for electromagnetic materials characterisation, namely electric conductivity, magnetic permeability, thermal conductivity and thermoelectric power. However, most measurements are also dependent on geometrical parameters and therefore yield extensive properties such as electric conductance, magnetic conductance and thermal conductance. It is interesting to point out that the thermoelectric power is an inherently intensive property, therefore it can be measured directly without geometric corrections, regardless of the size and shape of the material.

Because of the obvious discrepancy between the small number of feasible independent physical measurements and the potentially much larger number of uncertain material’s variables, the resulting MSA remains necessarily ambiguous. Gibbs’ well-known Phase Rule for homogeneous materials states that[3]:

F = C ! P + 2 ....................................(1)where F denotes the degrees of freedom, ie the number of independent intensive properties one must measure to fully characterise the material, C is the number of components, P is the number of phases in thermodynamic equilibrium with each other, and +2 stands for the two fundamental state variables of pressure and temperature. Recently, Olson et al suggested the following Microstructural Rule for intact materials based on Gibbs’ Phase Rule:

! = C " P + M + 2 ................................(2)where φ denotes the number of independent physical properties, ie the minimum number of independent measurements one needs to conduct to fully characterise the material, and M is the number of microstructural constituents or different phase morphologies[4]. The full thermo-mechanical history of the material exerts a convoluted effect on the material’s present microstructure, which leads to an inflation of the number of independent physical measurements needed to fully characterise the material. Although some of these changes can be classified as neutral microstructural evolution that exerts no significant influence on the remaining service life of the

material, they might very well exert a spurious influence on the measurable physical properties, thereby presenting a formidable obstacle for assessing those changes that can be classified as damage accumulation since they actually reduce the remaining service life of the material.

One important prediction of the Microstructural Rule given by Equation (2) is that reversible temperature and pressure effects on the measured physical property must be carefully taken into consideration even when irreversible effects, for example increasing dislocation density, precipitate coarsening, changes in solute partitioning, etc, are negligible. As an example, Figure 3 shows the electric resistivity versus temperature and pressure in pure nickel after Sundqvist[5]. ρP and ρV represent the electric resistivity at constant pressure and constant volume, respectively, so the difference between the two curves is due to the compressive stress caused by inhibited thermal expansion when the material is kept at constant volume. The behaviour is complicated by the fact that at room temperature nickel is ferromagnetic with a negative magnetostriction of around -30 ppm. Above the Curie temperature of TC ≈ 360ºC, the negative magnetostriction caused by spontaneous magnetisation and the resulting compressive stress at constant volume both disappear. The temperature coefficient of the electric resistivity is defined as follows:

!(T ) =" ln#

"T ....................................(3)

where ρ is the electric resistivity and T denotes temperature. The reversible temperature effect exhibits a slightly quadratic behaviour below TC with an average linear temperature coefficient of β ≈ 9.4 × 10-3 ºC-1, ie as little as 1ºC uncertainty in temperature could cause as much as 1% error in the measured electric resistance.

The pressure coefficient of the electric resistivity is defined as follows:

! (p) =" ln#

"p ....................................(4)

Figure 2. Complex relationships between material and component state parameters and measurable electromagnetic physical properties

Figure 3. Electric resistivity versus temperature and pressure in pure nickel after Sundqvist(5)

62 Insight Vol 52 No 2 February 2010

Insight Vol 52 No 2 February 2010 63

where p is the hydrostatic pressure (generally the electric resistivity is a second-order tensor that becomes anisotropic even in otherwise isotropic materials in the presence of directional deformation). At room temperature, the pressure coefficient in nickel is γ ≈ –1.9 × 10-2 GPa-1.

MSA monitoring faces a number of limitations including low sensitivity, lack of selectivity, lack of uniqueness, separation of intensive versus extensive properties, separation of reversible versus irreversible changes, inversion in the presence of non-monotonic changes, inter-dependence of various material changes, dominance of certain irrelevant microstructural changes that have no significant influence on remaining service life, etc. These limitations will be discussed separately in the following sections.

2.1 SensitivityUsually, MSA sensing techniques exhibit rather low sensitivity to subtle changes in the materials’ microstructure that nonetheless exert significant influence on their remaining service life. For example, increasing dislocation density and slip band formation in fatigue damaged specimens, changing precipitate density and size during thermal hardening, microscopic cavity formation at grain boundaries during creep damage accumulation, changing residual stress during thermo-mechanical relaxation, etc, all produce changes of a few percent or less in electric conductivity, thermoelectric power, and even less in thermal conductivity. Magnetic properties are typically more sensitive, but also less stable, therefore present similar problems in terms of relative sensitivity normalised to inherent material variations and measurement uncertainties.

2.2 SelectivityThe number of materials’ state variables of interest (microstructure, phase transformation, plastic strain, elastic strain, hardening, embrittlement, creep damage, fatigue damage, etc) is generally much larger than the number of independent physical measurements one can conduct in order to characterise the state of the material. Because of this fundamental limitation, the main problem with MSA is expected to be the lack of selectivity of any given measurement method rather than its potentially low sensitivity. For example, in most scenarios, the best residual stress monitoring method is not necessarily the one most sensitive to changes in the state of residual stress, but the one that is least sensitive to other spurious variations such as plastic strain, hardness, etc.

2.3 UniquenessLack of uniqueness in MSA could arise from two fundamental causes. The first cause is non-monotonic dependence of the measured physical property on the sought material state parameter. One example of this is illustrated in Figure 4, showing the electric conductivity versus Rockwell C Hardness in precipitation hardened IN718 nickel-base superalloy. As the hardness increases, the electric conductivity first drops from ≈1.56 %IACS at HRC 25 to ≈1.54 %IACS at HRC 32, then increases to ≈1.67 %IACS at HRC 46. Generally, the electric conductivity is a non-monotonic, therefore not invertible, function of hardness. However, over the hardness range of practical interest for engine applications (above HRC 40) the electric conductivity is a monotonic, therefore invertible, function of hardness, which could be exploited for MSA monitoring.

A second major cause of lack of uniqueness in MSA monitoring arises from the previously mentioned lack of selectivity in cases when the measured physical property changes by comparable amounts in response to more than one type of change in the state of the material. For example, the thermoelectric method is highly sensitive to subtle variation in material properties and the thermoelectric power (TEP) of ferritic steels can be used to monitor environmentally-assisted ageing in nuclear reactor plant components. Unfortunately, the selectivity of this method leaves

much to be desired, as illustrated by the following example. Figure 5 shows TEP (a) and hardness (b) as functions of temperature for 1-hour exposure time in JRQ steel[6]. Although thermal softening alone could be readily detected from the significant reduction of the thermoelectric power, parallel neutron radiation causes embrittlement accompanied by increased thermoelectric power. Figure 6 shows the changing thermoelectric power (a) and ductile-brittle transition temperature (b) during radiation embrittlement as functions of fluence for >1 MeV neutron irradiation in JRQ steel after Niffenegger et al[7]. These two effects could be both detected in the absence of the other, but can be separated when they occur simultaneously only by multi-modal inspection, for example by the additional monitoring of electric conductivity.

Monitoring of thermo-mechanical relaxation of residual stresses in surface-treated jet engine components offers another good example of lack of uniqueness. Residual stress assessment is important because there is mounting evidence that it is not possible to reliably and accurately predict the remaining service life of such components without properly accounting for the presence of residual stresses. Eddy current conductivity spectroscopy has emerged as the most promising approach for residual stress monitoring in surface-treated precipitation hardened nickel-base superalloys[8-26]. Unfortunately, cold work exerts a rather convoluted effect on

Figure 4. Electric conductivity versus Rockwell C hardness in precipitation hardened IN718 nickel-base superalloy

Figure 5. Thermoelectric power (a) and Vickers hardness (b) as functions of temperature for 1-hour exposure time in JRQ steel after Sevini et al(6)

residual stress profiling in shot-peened components by eddy current spectroscopic measurements. The spurious influence of cold work on eddy current residual stress assessment is much lower in specimens that are either laser shock-peened (LSP) or low-plasticity burnished (LPB), which exhibit much lower plastic deformation and therefore better thermo-mechanical stability than shot-peened ones. As an example, Figure 7 shows the comparison between eddy current and XRD results in low-plasticity burnished Waspaloy of approximately 15% maximum plastic strain after Abu-Nabah et al[24]. There is a fairly good agreement between the non-destructive eddy current and destructive XRD residual stress profiles because the low level of cold work induced by LPB. Similar agreement was also obtained for low-peening levels in powder metallurgic IN100 nickel-base superalloy in which the electric conductivity exhibits negligible sensitivity to cold work[24].

In general, compressive residual stress (elastic strain) below the mechanically treated surface results in increased electric conductivity, while parallel cold work (plastic strain) might either increase or decrease the conductivity depending on the hardness of the material. Figure 8 presents a clear example of how this combination of competing effects might lead to ambiguity in some cases. Under unfavorable conditions, which occur around HRC 40 in IN718, the two effects completely cancel each other in shot-peened components and even the best correction methods will necessarily fail. Figure 8 shows the apparent eddy current conductivity change (ΔAECC) spectra in twelve shot-peened specimens of Almen 4A

and 6A intensity. At around HRC 39 the expected positive ΔAECC changes sign, while above HRC 43 the negative ΔAECC is small but measurable. The continuous transformation of the regular positive ΔAECC spectrum into an irregular negative ΔAECC spectrum as the hardness increases indicates the stronger role of spurious cold work effects over the principal residual stress effect.

2.4 Intensive versus extensive propertiesSeparation of intensive properties from extensive ones is a major limitation of MSA monitoring. Most measurements are sensitive to both intensive (physical) properties of the material and extensive (geometrical) properties of the component. This important issue will be illustrated through two examples below. First, let us consider the feasibility of electric resistance measurement by four-point direct-current potential drop (DCPD) or low-frequency alternating-current potential drop (ACPD) techniques, which are uniquely suitable for monitoring slight wall thickness variations caused by corrosion and erosion. However, the ultimate limit of this extensive property monitoring is set by inherent variations in the underlying intensive property, ie electric resistivity, due to inevitable minor temperature variations.

Figure 9 shows the measured electric resistance variations in a ½"-thick 304 austenitic stainless steel plate during ACPD erosion monitoring over a period of 20 days. The measurement was done at 16 Hz where the standard penetration depth δ > 100 mm, ie much larger than the plate thickness. The separation between the sensing electrodes was 2" while the separation between the injection electrodes was 3". The electrodes were made of thermocouple wires in order to simultaneously monitor the plate’s temperature between ACPD measurements without injection. The reproducibility of the resistance measurement is around +/-0.02%, which is more than sufficient for monitoring purposes. The system

Figure 6. Thermoelectric power (a) and ductile-brittle transition temperature change (b) as functions of fluence during neutron radiation embrittlement in JRQ steel after Niffenegger et al(7)

Figure 7. The measured ΔAECC spectrum (a), cold work profile (b) and residual stress profile (c) in low-plasticity burnished Waspaloy after Abu-Nabah et al(24)

Figure 8. Apparent eddy current conductivity change (ΔAECC) spectra in twelve shot-peened specimens of Almen 4A and 6A intensity

64 Insight Vol 52 No 2 February 2010

Insight Vol 52 No 2 February 2010 65

was set up to ‘learn’ the temperature coefficient of the electric resistivity from the data as it accumulates. In order to reduce the inevitable variations of electric resistivity, we can exploit the fact that the underlying spurious temperature variations are oscillatory in nature without a statistically significant long-term trend, while the wall thickness changes monotonically and therefore exhibits a significant long-term trend. As it is illustrated in Figure 9, even well-defined distinct erosion events are difficult to detect without compensating for temperature oscillations that are merely a few ºC from peak to peak.

In the second example, thermoelectric inspection will be used to detect material property changes caused by plastic surface deformation. Although potential drop or eddy current conductivity measurements are also sensitive enough to detect the resulting conductivity variations, they are too sensitive to spurious variations in surface topography and changes in the extensive properties simply overshadow the sought changes in intensive properties[27-29]. It is well known that thermoelectric measurements are uniquely sensitive to changes in intensive material properties while being completely insensitive to extensive properties and, therefore, offer unique advantages over other methods in cases when discrimination between these two different classes of variables is necessary.

Self-referencing thermoelectric measurements can be done in an entirely non-contact way by using high-sensitivity magnetic detectors to sense the weak thermoelectric currents around inclusions and other types of inhomogeneity when the specimen to be tested is subjected to directional heating or cooling[30-34]. Figure 10 shows a schematic representation (a) of the closed-circuit Seebeck effects

and its typical realisation (b) in thermoelectric power measurements. External heating or cooling is applied to the specimen to produce a substantial temperature gradient in the region to be tested. As a result, different points on the boundary between the host material and the inclusion will be at different temperatures, therefore also at different thermoelectric potentials. These potential differences will produce opposite thermoelectric currents inside and outside the inclusion. The thermoelectric current forms local loops that run in opposite directions on the two sides of the inclusion relative to the prevailing heat flux, as it is indicated in Figure 10(b). When the specimen is scanned with a sensitive magnetometer, such as a fluxgate or Superconducting Quantum Interference Device (SQUID) detector, the magnetic field of these thermoelectric currents can be detected even if the inclusion is buried below the surface and the sensor is far away from the specimen.

The non-contacting thermoelectric technique is very sensitive to the presence of foreign body inclusions, when the thermoelectric power of the affected region is significantly different from that of the surrounding medium. The question arises whether mere plastic deformation of the material can produce a perceivable thermoelectric contrast with respect to the surrounding intact host. To answer this question, Figure 11 compares the magnetic scans of two apparently similar 9.53 mm-diameter surface holes in copper[35]. The left side of Figure 11 corresponds to a semi-spherical hole produced by low-stress milling which is expected to generate only negligible hardening and residual stress below the machined surface. In this case, no detectable magnetic field was observed from a 2 mm lift-off distance at ∇T ≈ 0.05°C/mm temperature gradient. In comparison, the right side of Figure 11 shows results from an otherwise similar semi-spherical indentation produced by pressing a stainless steel ball into the material in a manner that simulates a single impact during shot peening. As a result of plastic deformation, the surrounding material below and around the indentation is substantially hardened and supports significant residual stresses. The resulting peak magnetic flux was measured at ~36 nTpp, which is approximately 60 dB above the noise-limited detection threshold of the fluxgate magnetometer used in these experiments.

Figure 9. Electric resistance variations in a ½"-thick 304 austenitic stainless steel plate during ACPD erosion monitoring over a period of 20 days

Figure 10. Schematic representation (a) of the closed-circuit Seebeck effects and its typical realisation (b) in thermoelectric power measurements

The main concern in connection with shot-peened surfaces is not necessarily the presence, degree and penetration depth of the produced residual stress directly after surface treatment, but rather whether the stress has been significantly reduced or even eliminated by thermally-induced stress release after a long time spent in service. In order to demonstrate that the proposed thermoelectric method could readily detect this change, the bottom scans on Figure 11 show the magnetic images of the same two specimens after annealing them in a vacuum furnace for 30 minutes at 700°C. All the effects of plastic deformation during indentation, as well as the much weaker manufacturing texture found in the original bar stock, are gone. As a result, the thermoelectric currents are also eliminated and the measured magnetic field is essentially zero.

These results well demonstrate that the thermoelectric method is entirely insensitive to extensive parameters, such as surface topography, and therefore uniquely suitable for non-destructive assessment of subtle changes in intensive properties caused by plastic flow and residual stress formation. Whether the observed thermoelectric contrast is mainly due to the presence of residual stresses or to accompanying material effects of cold work depends on the type of material involved. In relative terms, the stress-induced variations of the thermoelectric power are similar to those of the ultrasonic velocity and electrical conductivity as they all range around 1% when the stress approaches the yield strength of the material. However, the thermoelectric power can be easily measured without any interference from geometrical variations, therefore lends itself much more readily to non-destructive assessment of near-surface residual stresses in shot-peened metals.

2.5 Reversible versus irreversible changesIn numerous cases, MSA monitoring requires the detection and quantitative assessment of relatively small irreversible changes in the presence of far larger reversible ones. The sought irreversible changes could be caused by microstructure evolution, phase transformation, precipitate formation and coarsening, microscopic cavity formation and coalescence, etc. Thermally-activated

processes often follow the Larson-Miller rule, originally developed for creep processes, that postulates that damage accumulation is proportional to the product of absolute temperature and a time-dependent term that is linearly proportional to the logarithm of exposure time[36]. The separation problem arises from the fact that, in terms of a given physical parameter, for example electric resistivity, the slowly accumulating irreversible change is typically only a few percents while the reversible temperature-dependence of the same parameter could be much higher, in some cases as high as 100% or even more.

2.6 Inter-dependenceMSA monitoring is further complicated by strong inter-dependence between thermal exposure and pre-existing microstructural damage. Thermally-activated microstructural evolution and residual stress relaxation both start at a significantly lower temperature and progress at a significantly faster pace in the presence of microstructural damage, for example plastic deformation due to cold work. Following such processes through changes in any given physical property is adversely affected by the fact that the measured parameter is not only sensitive to direct effects of the microstructural damage, but also to its indirect effect of influencing the thermally activated change in other materials’ state parameters.

2.7 DominanceDominance is the phenomenon of overshadowing of all significant changes in the material that could be exploited for MSA monitoring by a single microstructural effect that has no perceivable effect on the remaining service life of the material. For example, low-cycle fatigue damage monitoring in austenitic stainless steel components is of primary importance in certain fracture-critical nuclear reactor plant components. Numerous magnetic, electric, ultrasonic and thermoelectric methods could be exploited for this purpose except that all of them are excessively sensitive and therefore overwhelmed by martensite formation that is not directly related to the resulting life reduction.

The main problem is that this dominant effect is strongly temperature dependent, as illustrated in Figure 12 which shows the level of martensite formation as a function of temperature due to low-cycle fatigue and quasi-static plastic deformation after Kalkhof et al[37]. Depending on whether otherwise negligible plasticity took place at 25ºC or 40ºC, the amount of generated martensite could change by as much as 100%, thereby changing most measurable physical parameters, especially those affected by magnetic properties. Figure 13 shows the change in magnetic susceptibility and apparent eddy current conductivity (AECC) caused by 15% plastic deformation at different temperatures. The typical operational temperature of components made of this material would be around 300ºC where no significant martensite formation could occur. Although it is possible to detect very low levels of martensite formation even at such temperatures using

Figure 12. Martensite formation as a function of temperature due to low-cycle fatigue and quasi-static plastic deformation after Kalkhof et al(7)

Figure 11. Non-contact thermoelectric detection of plastic flow caused by surface indentation in copper after Carreon et al(35)

66 Insight Vol 52 No 2 February 2010

Insight Vol 52 No 2 February 2010 67

sensitive magnetometers such as SQUIDs, the measured quantity would be most likely dominated by spurious room-temperature artifacts produced accidentally during construction and regular maintenance.

In the following sections of this article, two MSA monitoring techniques that are especially well suited for in-situ continuous MSA monitoring will be described in more detail to illustrate some of the above listed inherent limitations. The first example will be thermoelectric monitoring, followed by low-frequency alternating current potential drop measurement.

3. In-situ thermoelectric monitoringThermoelectricity is caused by coupled transport of heat and electricity in metals and semiconductors that leads to a number of interesting phenomena, some of which can be exploited for NDE and materials characterisation. The coupled constitutive equations can be written as follows:

Jh

!

"#

$

%& =

' 'S

'ST (

!

"#

$

%&)*V

)*T

!

"#

$

%&

........................(5)

where J is the electric current density, h is the thermal flux density, V is the electric potential, T denotes temperature, σ is the electric conductivity at ∇T = 0, κ is the thermal conductivity at ∇V = 0, and S is the thermoelectric power. Most existing thermoelectric NDE methods are based on the well-known Seebeck effect that is commonly used in thermocouples to measure temperature at the junction between two different conductors.

Figure 14 shows the schematic representation (a) of the open-circuit Seebeck effect and its typical realisation (b) in thermoelectric NDE. In this commonly used configuration the current density J is always zero, therefore Equation (5) yields ∇V = – S∇T. One of the reference electrodes is heated by electrical means to a preset temperature of 100-300ºC, pretty much like the tip of a temperature-stabilised soldering iron, and connected to the inverting (–) input of the differential amplifier driving the indicator. The other electrode is left cold at essentially room temperature and connected to the non-inverting (+) input. The measurement is done quickly enough to assure (i) that the hot reference electrode is not cooled down perceivably by the specimen and (ii) that the rest of the specimen beyond the close vicinity of the contact point is not warmed up perceivably. Assuming that the relative thermoelectric power of the specimen with respect to the reference electrode SAB = SA – SB does not change significantly between the temperatures of the hot T2 and cold T1 junctions, the measured thermoelectric voltage can be calculated as follows:

V = (SA! S

B)dT " S

AB(T

2! T

1)

T1

T2

# ....................(6)

Ideally, regardless of how high the temperature difference between the junctions is, only thermocouples made of different materials, ie materials of different thermoelectric powers, will generate thermoelectric signals. This unique feature makes the simple thermoelectric tester one of the most sensitive material discriminators used in non-destructive inspection. The thermoelectric power of metals is sensitive to a variety of material properties that can affect the measurement. Clearly, chemical composition exerts the strongest effect on the thermoelectric properties and, accordingly, the basic application of thermoelectric materials characterisation is metal sorting. However, it is known that materials of identical chemical composition can also produce an efficient thermocouple as a result of different heat treatments, hardening, texture, fatigue, etc, which can be further exploited for non-destructive testing of materials[38-45].

High-precision continuous in-situ TEP measurements require

Figure 13. Change in magnetic susceptibility (a) and apparent eddy current conductivity (AECC) (b) caused by 15% plastic deformation in SS304 at different temperatures

Figure 14. Schematic representation (a) of the open-circuit Seebeck effects and its typical realisation (b) in thermoelectric NDE

accurate knowledge of the junction temperatures, which can be most easily achieved using the four-wire, ie two-pair, configuration shown in Figure 15[46-49]. In this case, two temperature and two voltage readings are taken in each measurement cycle using the C1-A1 (T1), C2-A2 (T2), A1-A2 (V3) and C1-C2 (V4) combinations of the thermocouple wires (C and A stands for the Chromel and Alumel, respectively, wires of a K-type thermocouple). Initially, the specimen is in thermal equilibrium with a uniform ambient temperature. Then, the specimen is exposed to localised heating so that the first junction (C1-A1) is heated to T1 while the second junction (C2-A2) is heated to T2, which is either lower or higher than T1 depending on where the heating is applied. Figure 16 shows an example of the temperature (a) and voltage (b) waveforms produced by a ferritic steel specimen using a two-pair K-type thermocouple electrode sensor. The absolute thermoelectric power of the specimen can be calculated as follows:

SS =1

2

V3+V

4

T2! T

1

+SA + SC

2 ............................(7)

where the bar indicates averaging of the temperature-dependent TEP between T1 and T2. The actually measured relative TEP is the difference between the average TEP of the specimen SS and the combined average TEP of the two thermocouple wires (SA + SC ) / 2 . Since K-type thermocouples can be used up to 1200ºC without time constraint, it is assumed that this average TEP (about +2 μV/ºC at RT) does not change irreversibly during thermal exposure, though it certainly exhibits some reversible temperature dependence.

Figure 17(a) shows the measured relative TEP and average temperature during a room-temperature test lasting almost 12 days. It is clear that there exists a correlation between the observed TEP variation and the oscillation of the average temperature, which follows the minute daily oscillations of the ambient temperature in the otherwise air-conditioned laboratory. This effect can be better illustrated by plotting the TEP as a function of the average temperature, as it is done in Figure 17(b). A clear linear correlation can be observed between the measured TEP and the average temperature even over this very small range of ~3ºC temperature variation in the air-conditioned laboratory. This correlation indicates the strong temperature-dependence of the TEP and is in line with expectations based on typical values listed in the literature. Figure 18 shows the results of an extended test of three different ferritic steels (JRQ, 508 and 533) for 38 days of total exposure time over 76 days. During the first 26 days of this test, the specimens were exposed to 410ºC for 12 hours each day, then the ageing temperature was increased to 510ºC. The fourth specimen (JRQ-JB ref) was kept at room temperature throughout the whole time to monitor the long-term drift of the measurement system itself. It appears that the changes in TEP caused by 12-hour daily exposure to 510ºC stabilise after about 50 days, ie after ~25 days of total exposure in Figure 18.

4. In-situ potential drop monitoringElectrical resistivity measurements with potential drop methods can be exploited for continuous in-situ MSA monitoring. For example,

Figure 15. A schematic diagram of two-pair temperature/thermoelectric power measurement

Figure 16. Temperature (a) and voltage (b) waveforms produced by a ferritic steel specimen using a two-pair K-type thermocouple electrode sensor

Figure 17. Measured relative TEP and average temperature during a room-temperature test (a) and the same data plotted as a function of temperature (b)

68 Insight Vol 52 No 2 February 2010

Insight Vol 52 No 2 February 2010 69

thermal exposure[63]. Continuous in-situ resistivity monitoring was conducted throughout various heating cycles on fully precipitation hardened (VH 412) IN-718 nickel-base superalloy specimens with different levels of cold work as they were exposed to gradually increasing peak temperatures from 400°C to 800°C. The electrode wires were spot welded to the specimens. Inevitable positioning errors could cause as much as ±10-15% variation in the measured absolute resistance. Therefore, the measured resistance changes were normalised to the initial resistance of the specimen.

Figure 19 shows the electric resistivity variation history in a mechanically intact IN718 nickel-base superalloy specimen (a) and a cold-worked one of 30% plastic strain (b). The initial electric conductivities of the intact and cold-worked specimens were 1.62 and 1.57 %IACS, respectively. Both specimens were subjected to nine cycles of one hour each at peak temperatures increased from 400°C to 800°C in steps of 50°C. The results indicate a monotonic increase in resistance at the peak temperatures for the intact specimen. For the 30% compressed sample, the resistance at peak temperature initially increases until the peak temperature reaches 600°C and decreases thereafter. It is important to point out that the irreversible resistance variations in the 30% compressed specimen are about one order of magnitude higher than in the intact one. Moreover, the onset of the microstructural evolution is at least 50°C earlier in the cold-worked sample than in the intact sample, ie the microstructural evolution starts not only at lower temperature, but also progresses at a faster rate in the presence of cold work. Figure 20 shows the electric resistivity variation versus temperature in mechanically intact (a) and cold-worked to 30% plastic strain (b) IN718 nickel-based superalloy specimens. As expected, cold work not only increased the electric resistivity by about 3%, it also decreased the temperature coefficient of the resistivity from ~280 ppm/°C to ~200 ppm/°C, a significant drop of almost 30%. However, above 700°C the original resistivity decreases as the microstructural damage is repaired and the temperature coefficient reverts to its original high value of ~280 ppm/°C.

Figure 21 shows the irreversible electric resistivity change as a function of cold work in IN718 after different exposure times at 550ºC (a) and 650ºC (b). The measured resistance values were normalised to the corresponding resistance of the specimen before

this technique has been used for monitoring thermal exposure in steel components[50-52]. Several papers have reported that long-term thermal exposure of low-alloy ferritic steels causes a reduction of the electrical resistivity and that applied stresses lead to even greater reductions[53-57]. Under creep loading, part of the increased resistivity is due to increasing porosity, but microstructural evolution also contributes to the change in resistivity. All studies suggest that electrical resistivity monitoring can be used for the measurement of volumetric creep damage, especially at early stages when creep damage is hardly detectable by other methods[58-61].

As it was mentioned earlier, MSA monitoring often requires the detection and quantitative assessment of relatively small irreversible changes in the presence of far larger reversible changes. The slowly accumulating irreversible effect typically changes the electric resistivity by only a few percent, while the reversible change over the operational temperature range (a few hundred ºC) could be well above 100%. It is important to recognise that the reversible temperature effect is essentially isotropic as increasing lattice vibrations at elevated temperature increase the resistivity by the same amount in every direction. As opposed to the often-used in-line electrode configuration, the rectangular arrangement shown earlier in Figure 15, especially the square configuration, can be used as a directional sensor to monitor separately the electric resistivity in two orthogonal directions without heating[62]. This method can separate the common (isotropic) part of the change from the differential (anisotropic) part. The reversible temperature effect is entirely isotropic, while the irreversible temperature effect is mostly isotropic but might have a perceivable anisotropic contribution in the case of pre-existing crystallographic or morphological texture in the microstructure. Usually, the thermo-mechanical effect, ie creep damage, is also a combination of isotropic and anisotropic contributions that can be separated using directional electrode configuration by comparing the common and differential parts of the resistance change. Simple DCPD or low-frequency ACPD electric sensors work well for both reversible elastic strains on the order of 10-3 and irreversible plastic strains well over 10-1, and have been successfully tested up to well over 1000ºC.

It was mentioned earlier that thermally-activated microstructural evolution and residual stress relaxation both start at significantly lower temperature and progress at significantly faster pace in the presence of microstructural damage, for example plastic deformation due to cold work. Following such processes through changes in any given physical property, such as electric resistivity, is adversely affected by the fact that the measured parameter is not only sensitive to direct effects of the microstructural damage, but also to its indirect effect of influencing the thermally activated change in other materials’ state parameters.

In order to illustrate the significant influence of pre-existing microstructural damage on the initiation and progress of thermally activated microstructural changes, let us consider the influence of cold work on low-frequency ACPD resistivity variations with

Figure 18. Thermoelectric monitoring of a 75-day thermal ageing test of different ferritic steels

Figure 19. Electric resistivity variation history in mechanically intact (a) and cold-worked to 30% plastic strain (b) IN718 nickel-base superalloy specimens

thermal exposure in order to avoid the effect of variations in initial microstructure and electrode positioning. Both peak temperatures show that the magnitude of the resistance variation increases with cold work and the microstructure evolution occurs at a notably faster rate with increased cold work. The initial irreversible rise in resistivity is approximately one order of magnitude higher and occurs at about 50°C lower temperature in cold-worked samples of 30% plastic strain than in the intact material.

5. ConclusionsSignificant reduction in the ability of a fracture-critical component to perform its function generally begins at the material level, therefore successful implementation of condition-based maintenance hinges on full characterisation of the current state of material, which is referred to as materials’ state awareness or MSA. The main technology challenges associated with this multidisciplinary concept are (i) to assess early and progressive changes in material state associated with operational usage and exposure, (ii) to predict the real-time physical, chemical or electronic state for complex systems subject service loads and environmental exposure over time, and (iii) to relate the current and evolving state of microstructure and damage processes to enable probabilistic prognosis modelling of the material/structural/system state. Inspection methods for MSA will have to be based on simple, reliable and rugged non-destructive monitoring procedures that can be easily adapted to field conditions. This paper reviewed some of the most popular electric and thermoelectric non-destructive evaluation methods that might be applicable for materials’ state awareness monitoring. A variety of sensors can be built based on these principles to detect and quantitatively characterise subtle environmentally-assisted and/or service-related changes in metals.

MSA monitoring faces a number of limitations that include

low sensitivity, lack of selectivity, lack of uniqueness, separation of intensive versus extensive properties, separation of reversible versus irreversible changes, inversion in the presence of non-monotonic changes, inter-dependence of various material changes, dominance of certain irrelevant microstructural changes that have no significant influence on remaining service life, and so on. The most important conclusion one can draw from relevant studies published in the scientific literature is that, in most cases, the detection sensitivity is sufficiently high for the purposes of materials’ state awareness monitoring and the feasibility of the sensing method is mainly determined by its selectivity, or the lack of it, to a particular type of damage mechanism. It is unfortunate that far too many studies are focused on sensitivity alone to prove feasibility rather than addressing the critical issue of selectivity which is a more important feature of any materials’ state monitoring technique than sensitivity itself.

In order to illustrate some of the above-listed inherent limitations, this paper reviewed two MSA sensing techniques that are especially well suited for in-situ continuous monitoring, namely the thermoelectric and low-frequency alternating current potential drop techniques. In both cases the sensor consists of the material to be tested and two pairs of thermocouple wires attached to it, ie there is no need for transducers that convert one type of a physical parameter into another. One unique feature of this simple and rugged four-point electrode configuration is that it can be exploited for both thermoelectric and electric resistivity monitoring of the material during service including thermo-mechanical and radiation exposure. This is particularly important in cases when more than one type of damage accumulation and/or microstructural evolution takes place during service because separation of these simultaneous changes in the state of the material requires a multi-modal monitoring approach.

Figure 20. Electric resistivity variation versus temperature in mechanically intact (a) and cold-worked to 30% plastic strain (b) IN718 nickel-base superalloy specimens Figure 21. Irreversible electric resistivity change as a function

of cold work in IN718 after different exposure times at 550ºC (a) and 650ºC (b)

70 Insight Vol 52 No 2 February 2010

Insight Vol 52 No 2 February 2010 71

AcknowledgementsThe author would like to acknowledge valuable contributions to the presented results by Curtis Fox, Elise Minda and Elhoucine Madhi of the University of Cincinnati, Giuseppe Sposito and Peter Cawley of Imperial College London, Bassam Abu-Nabah of General Electric Aviation, Mark Blodgett of the Air Force Research Laboratory, and Triplicane Parthasarathy of Universal Energy Systems, Inc.

References1. Materials State Awareness, National Materials Advisory Board,

National Research Council of the National Academies, ed E N Meyer, The National Academies Press, Washington, 2008.

2. Prognosis of Aircraft and Space Devices, Components and Systems, Discovery Challenge Thrust, AFOSR, 2008.

3. J W Gibbs, The Scientific Papers of J Willard Gibbs, Vol 1, p 96, Dover, New York, 1961.

4. D L Olson, V I Kaydanov, A N Lasseigne-Jackson, A Landau, and M Pinkas, in Review of Progress in Quantitative NDE, Vol 27, pp 3-22, American Institute of Physics, Melville, 2008.

5. B Sundqvist, Phys Rev B, 38, 12283-12289, 1988.6. F Sevini, B Acosta and L Debarberis, Int J Press Vess Piping,

83, 525-530, 2006.7. M Niffenegger, K Reichlin and D Kalkhof, Nucl Eng Design,

235, 1777-1788, 2005.8. F C Schoenig Jr, J A Soules, H Chang and J J DiCillo, Materials

Evaluation, 53, 22-26, 1995.9. M Blaszkiewicz, L Albertin and W Junker, Mater Sci Forum,

210, 179-186, 1996.10. H Chang, F C Schoenig Jr and J A Soules, Materials Evaluation,

57, 1257-1260, 1999.11. I Lavrentyev, P A Stucky and W A Veronesi, in Review of

Progress in Quantitative NDE, Vol 19, pp 1621-1628, American Institute of Physics, Melville, 2000.

12. V Zilberstein, Y Sheiretov, A Washabaugh, Y Chen and N Goldfine, in Review of Progress in Quantitative NDE, Vol 20, pp 985-995, American Institute of Physics, Melville, 2001.

13. V Zilberstein, M Fisher, D Grundy, D Schlicker, V Tsukernik, V Vengrinovich, N Goldfine and T Yentzer, ASME J Press Vessel Techn, 124, 375-381, 2002.

14. M P Blodgett and P B Nagy, J Nondestr Eval, 23, 107-123, 2004.

15. F Yu and P B Nagy, J Appl Phys, 96, 1257-1266, 2004.16. F Yu and P B Nagy, J Nondestr Eval, 24, 143-152, 2005.17. F Yu and P B Nagy, J Nondestr Eval, 25, 107-122, 2006.18. F Yu, M P Blodgett and P B Nagy, J Nondestr Eval, 25, 17-28,

2006.19. N Nakagawa, C Lee and Y Shen, in Review of Progress in

Quantitative NDE, Vol 25, pp 1418-1425, American Institute of Physics, Melville, 2006.

20. B A Abu-Nabah and P B Nagy, NDT&E Int, 39, 641-651, 2006.21. B A Abu-Nabah and P B Nagy, NDT&E Int, 40, 405-418, 2007.22. B A Abu-Nabah and P B Nagy, NDT&E Int, 40, 555-565,

2007.23. Y Shen, C Lee, C C H Lo, N Nakagawa and A M Frishman, J

Appl Phys, 101, 014907, 2007.24. B A Abu-Nabah, F Yu, W T Hassan, M P Blodgett and P B

Nagy, Nondestr Testing Eval, 24, 209-232, 2009.25. S Hillmann, H Heuer, H U Baron, J Bamberg, A Yashan and N

Meyendorf, in Review of Progress in Quantitative NDE, Vol 28, pp 1349-1355, American Institute of Physics, Melville, 2009.

26. Y Sekine and H Soyama, Surf Coat Techn, 203, 2254-2259, 2009.

27. M P Blodgett, C V Ukpabi and P B Nagy, Mater Eval, 61, 765-772, 2003.

28. K Kalyanasundaram and P B Nagy, NDT&E Int, 37, 47-56, 2004.

29. F Yu and P B Nagy, J Appl Phys, 95, 8340-8351, 2004.30. J H Hinken and Y Tavrin, in Review of Progress in Quantitative

NDE, Vol 19, pp 2085-2092, American Institute of Physics, Melville, 2000.

31. P B Nagy and A H Nayfeh, J Appl Phys, 87, 7481-7490, 2000.32. H Carreon, P B Nagy and A H Nayfeh, J Appl Phys, 88, 6495-

6500, 2000.33. K Maslov and V K Kinra, Mater Eval, 59, 1081-1084, 2001.34. A H Nayfeh, H Carreon and P B Nagy, J Appl Phys, 91, 225-

231, 2002.35. H Carreon, P B Nagy and M P Blodgett, Res Nondestr Eval, 14,

59-80, 2002.36. F R Larson and J Miller, Trans ASME, 74, 765-775, 1952.37. D Kalkhof, M Grosse, M Niffenegger and H J Leber, Fatigue

Fract Eng Mater Struct, 27, 595-607, 2004.38. C M Stuart, J Testing Eval, 15, 224-230, 1987.39. W Morgner, Mat Eval 9, 1081-1088, 1991.40. J P Fulton, B Wincheski and M Namkung, in Review of

Progress in Quantitative NDE, Vol 13, pp 1611-1618, Plenum Press, New York, 1994.

41. E B Henry, C M Stuart and W Tomasulo, in Nondestructive Testing Handbook, ed P McIntire, Vol 9, pp 363-376, ASNT, Columbus, 1995.

42. J Hu and P B Nagy, Appl Phys Lett, 73, 467-469, 1998.43. P B Nagy and J Hu, in Review of Progress in Quantitative

NDE, Vol 17, pp 1573-1580, Plenum, New York, 1998.44. A W Szafranski, J Alloys Compd, 340, 54-61, 2002.45. P Termsuksawad, S Niyomsoan, R B Goldfarb, V I Kaydanov,

D L Olson, B Mishra and Z Gavra, J Alloys Compd, 373, 86-95, 2004.

46. E Nagy and J Toth, J Phys Chem, Solids 24, 1043-1047, 1963.47. R D Barnard, Thermoelectricity in Metals and Alloys, pp

38-39, Taylor & Francis, London, 1972.48. A K Singh and G Ramani, Rev Sci Instru, 49, 1324-1326,

1978.49. D A Polvani, J F Meng, M Hasegawa and J V Badding, Rev Sci

Instrum, 70, 3586-3589, 1999.50. M Sukekawa, M Shiga, Y Tobita and S Kirihara, in International

Conference on Creep, pp 321-325, Jap Soc Mech Eng, Tokyo, 1986.

51. B Dogan, K Nikbin, B Petrovski, U Ceyhan and D W Dean, Int J Press Vess Piping, 83, 784-797, 2006.

52. F Masuyama, Int J Press Vess Piping, 83, 819-825, 2006.53. S Kirihara, M Shiga, M Sukekawa, T Yoshioka and C J Asano,

Soc Mat Sci (Japan), 33, 1097-1102, 1984.54. M Sukegawa, M Shiga, C Asano, Y Tobita and S Kirihara, J

Soc Mat Sci (Japan), 34, 1478-1482, 1985.55. J Bressers, W Hessler, U W Hildebrandt and H Willems, in

High Temperature Materials for Power Engineering, Part 1, pp 629-642, Kluwer Academic Publishers, Norwell, 1990.

56. S H Nahm, K M Yu, J S Park, Y H Nam, S I Han, J M Kim and A K Kim, Int J Mod Phys, B 17, 1615-1620, 2003.

57. S H Nahm, J M Kim, J S Park, K M Yu, D K Kim, A Kim and D J Kim, Key Engg Mater, 270-273, 1212-1217, 2004.

58. H Willems and G Dobmann, Nucl Engg Design, 128, 139-149, 1991.

59. F Masuyama, T Tokunaga, N Shimohata, T Yamamoto and M Hirano, in Creep & Fracture in High Temperature Components: Design & Life Assessment Issues, pp 19-30, DEStech Publications, Lancaster, 2009.

60. R M Pelloux, J M Peltier and V A Zilberstein, J Engg Mater Techn, 111, 19-20, 1989.

61. K M Yu, S H Nahm and Y I Kim, J Mater Sci Lett, 18, 1175-1176, 1999.

62. V M Tatarnikov, Meas Techn, 13, 877-881, 1970.63. E Madhi and P B Nagy, in Review of Progress in Quantitative

NDE, Vol 28, Plenum, New York, to be published 2010.