Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Modeling the nonlinear hysteretic response

in DAE experiments of Berea sandstone: A


case-study
Cite as: AIP Conference Proceedings 1650, 1553 (2015); https://doi.org/10.1063/1.4914774
Published Online: 02 April 2015

Claudio Pecorari

AIP Conference Proceedings 1650, 1553 (2015); https://doi.org/10.1063/1.4914774 1650, 1553

© 2015 AIP Publishing LLC.


Modeling the Nonlinear Hysteretic Response in DAE
Experiments of Berea Sandstone: a Case-Study

Claudio Pecorari

Hesjakollen 111 A
5142 Bergen, Norway

claudio.pecorari@hotmail.com

Abstract. Dynamic acousto-elasticity (DAE) allows probing the instantaneous state of a material while the latter slowly
and periodically is changed by an external, dynamic source. In DAE investigations of geo-materials, hysteresis of the
material’s modulus defect displays intriguing features which have not yet been interpreted in terms of any specific
mechanism occurring at atomic or mesoscale. Here, experimental results on dry Berea sandstone, which is the rock type
best investigated by means of a DAE technique, are analyzed in terms of three rheological models providing simplified
representations of mechanisms involving dislocations interacting with point defects which are distributed along the
dislocations’ core or glide planes, and microcracks with finite stiffness in compression. Constitutive relations linking
macroscopic strain and stress are derived. From the latter, the modulus defect associated to each mechanism is recovered.
These models are employed to construct a composite one which is capable of reproducing several of the main features
observed in the experimental data. The limitations of the present approach and, possibly, of the current implementation of
DAE are discussed.

INTRODUCTION
If the mathematical description of the effect of damage on the mechanical properties of a material is all that is
needed, then suitable constitutive relationships suffice. On the other hand, if understanding the significance of a
particular set of experimental results in the context of structural safety is required, then the physical mechanisms
responsible for a specific form of the constitutive relationships must be identified. In what follows, an attempt at
interpreting recent experimental results on Berea sandstone in terms of specific microscopic mechanisms is presented.
To this end, constitutive relationships linking anelastic strain to the applied stress are introduced.
A material tested by the dynamic acousto-elastic technique (DAET) is subjected simultaneously to a low-
frequency, high amplitude wave, (a.k.a. the pump wave) which continuously changes the state of the material in time,
and to a high-frequency, low-amplitude pulse, which probes the material’s current state. The frequency of the pump
wave is selected to coincide with that of a resonant mode of the material sample and is usually of the order of a few
kHz. In view of the dynamic nature of the high-amplitude excitation, the sample is tested both in compression, as it is
usually done in more conventional configurations using quasi-static loading (see for instance [1]), and in tension. This
property of DAET allows access to important information about damage precursors by simply comparing the response
of the material to external stresses of opposite sign.
Renaud et al. [2] carried out an investigation on eleven rock types with different microstructure using DAET. The
diversity of microstructure was shown to be accompanied by an equally remarkable variety of dynamic acousto-elastic
response. Renaud et al. [3], and Rivière et al. [4] recently employed DAET to further investigate the nonlinear
response of dry Berea sandstone. In addition to the classical nonlinear elastic response, these results document
hysteretic dynamics displaying complexity which increases with the amplitude of the pump wave, and suggests that
more mechanisms may be simultaneously active.

41st Annual Review of Progress in Quantitative Nondestructive Evaluation


AIP Conf. Proc. 1650, 1553-1560 (2015); doi: 10.1063/1.4914774
© 2015 AIP Publishing LLC 978-0-7354-1292-7/$30.00

1553
The magnitude of nonlinear, hysteretic dynamics correlates with the severity of defects at atomic and/or meso-
scale. However, the role of these imperfections in determining the nonlinear character of the material’s response to an
inspecting wave remains still quite unclear. Several mechanisms at the atomic level and/or structural properties at
meso-scale (see for example [1, 5-8]) have been proposed to explain hysteresis of mechanical properties of materials
with damage. Soft bonds, micro-cracks, dislocations, fluid trapped within pores are among the structural features
most cited in this context. Specific mechanisms called upon involve friction, capillary forces, adhesion, residual
compressive stresses, and bond breaking and forming.
Voids, which may enhance the classical component of the material nonlinearity, but are unlike sources of hysteretic
phenomena are not considered here. Capillary forces are excluded from consideration since the samples investigated
in [3] and [4] are said to be dry. Friction forces between faces of partially closed micro-cracks may also be discarded.
In fact, a crack, which is subjected to a traction force with both tangential and normal component, does not dissipate
mechanical energy via friction forces when it is open. Similarly, when the component of the traction, which is normal
to the crack, becomes compressive, tangential motion is progressively hindered as the crack’s closure increases [9].
This behavior is in stark contrast with that displayed by the experimental results which show hysteretic effects
increasing with the amplitude of the excitation. Therefore, experimental results provide conclusive evidence excluding
friction forces between microcrack faces in relative motion as a relevant dissipative mechanism in dry Berea
sandstone.
The main candidates which are left as potential causes of the observed hysteresis in dry Berea sandstone are
dislocations interacting with point defects, sliding grain boundaries, and adhesion between partially closed microcrack
faces. To model an effective instantaneous stiffness, which is asymmetric with respect to the tension-compression
phases of the loading cycle, additional mechanisms like the dynamics of micro-cracks with finite stiffness, and
classical nonlinear elasticity, may be considered.
In this work, two effective constitutive relations are used to recover the main features of the anelastic, hysteretic
response of dry Berea sandstone. These constitutive relations are built from rheological models that have been
employed to describe the dynamics of dislocations interacting with point defects distributed in close proximity of their
glide planes (glide PDs), or with point defect located along their cores (core PDs). The constitutive relationship used
for the former system can be employed also to model friction forces between sliding grain boundaries. Indeed, the
latter may be modelled in terms of arrays of dislocations. The relationship employed to simulate the dislocation-core
PD interaction under tension qualitatively matches the one describing the anelastic effect caused by adhesion between
crack faces [10]. In compression, however, the latter mechanism is not activated because of crack closure, while
dislocation-core PDs continues to generate anelastic strain.
Two additional constitutive relationships are considered which do not account for dissipative effects. The first one
considers a distribution of micro-cracks exhibiting finite stiffness during closure, while, during tension, they are
assumed to be open and with stress-independent, finite compliance. The second one, of classical nature, prescribes a
quadratic dependence of the modulus defect on the applied stress, and is considered only at a later stage.
These four relations are used to build a composite, macroscopic model by means of which the normalized modulus
defect measured by DAET in dry Berea sandstone is recovered. Comparison between predictions of this model and
the experimental results in [3] and [4] are discussed together with aspects of the latter which are not accounted for by
the model presented here.

MODELING DAET
In this section, four constitutive relations are derived from which the corresponding relative modulus defect is
obtained as a function of the applied stress, V. The reader should not regard these models as attempts to describe the
defect’s dynamics in all its details. Rather, where it applies, the objective of these models is to capture the core of
distinct amplitude-dependent mechanisms by which the energy of a dynamic excitation is dissipated.

Dislocation – Glide Point Defect Interaction


This and the following sub-section focus on the interaction between dislocations and point defects, and borrow
from [1] the basic rheological model on which the constitutive relations of interest are eventually built. Point defects
located in close proximity to a glide plane of a dislocation interact with the latter in a way which is schematically
represented in the inset of Fig. 1. This diagram shows that once a critical value, Vcr, of the maximum applied stress,

1554
Vmax, is overcome, the strain develops an anelastic hysteretic component, ean. The hysteresis associated to this
interaction is characterized by a fully plastic response immediately after the two inversion points in the stress cycle.
Following this phase, ean displays a linear dependence on Vwhich can be viewed as direct result of the long-range
interaction between these defects.
The interaction of interest here acts over a long-range, and its intensity is controlled, among other factors, by the
density and physical nature of the point defects. For this reason, it seems reasonable to assume that the quantity Vcr is
not unique, but it is distributed according to a certain law. In this investigation, the following distribution is employed
for Vcr

Το
ɔ(ɐcr ) = (ɐcr Τοଶ )݁ ିɐ cr . (1)

The product M(Vcr)dVcr represents the fraction of dislocations breaking away from point defects when the applied
stress, V, is between Vcr and (Vcr + dVcr). The fraction, f, of dislocations which have broken away as a function of Vis
obtained by integrating M(Vcr) over [0, V] to obtain

f(ɐ)=1-൫1+(ɐΤο)൯e-ɐΤο . (2)

Using this information, the following model for the anelastic component of the strain, which is produced by the
interaction of dislocations with glide PDs, is obtained

ean =S1 {[ɐ-(ɐmax +ɐ)(1+z+ )e-z+ +ɐmax (1+zo )e-zo -


ʹο(1-2(1+z+ + z+2 Τ2)e-z+ +(1+zo + zo2 Τ2)e-zo )]H(ɐሶ )+ (3)
[ɐ+(ɐmax -ɐ)(1+z- )e-z- -ɐmax (1+zo )e-zo +
ʹο(1-2(1+z- + z-2 Τ2)e-z- +(1+zo + zo2 Τ2)e-zo )]H(െɐሶ )} ,

where z± = (ɐmax ±ɐ)Τ(ʹο), zo = ɐmax Το. Figure 1 provides a graphical representation of eq. (3).
By taking the derivative of the total strain with respect to V the relative modulus defect associate to eq. (3) is found
to be
οE / Eo ؆ -ɂ[H(ɐሶ )f(z+ )+H(-ɐሶ )f(z- )] . (4)
In Fig. 2, three examples of eq. (4) corresponding to values Vmax/' =1, 3, 5andH  S1/So) = 1, where So is the linear
compliance of the material, are shown.

8 -6 -4 -2 0 2 4 6
ean 0
6
Figure 1. Anelastic strain versus stress for a V'
distribution of dislocations
4 interacting with -0.2
point defects
Vcr Vmax
along their glide planes. The
inset illustrates
2
the rheological model -0.4
underlying the macroscopic
0 strain-stress
relationship-5of a material
-10 0 with dislocations.
5 V 10
-2 -0.6
-4
-0.8
-6
-8 -1 'E/Eo
FIGURE 1. Anelastic strain versus stress for a distribution of FIGURE 2. Relative modulus defect, 'E/Eo, versus V' for
dislocations interacting with glide PDs. The inset illustrates Vmax' = 1, 3, and 5, caused by dislocations interacting with
the rheological model underlying the macroscopic strain- glide PDs. This figure assumes H = 1 in eq. (4).
stress relationship of a material with dislocations interacting
with glide PDs.

1555
Dislocation – Core Point Defect Interaction
Point defects can be found also within the core region of dislocations and are responsible for a short-range
interaction with the latter. In the linear regime, the effect of dislocation-core PD interaction on the propagation of
ultrasonic waves, and in particular on wave attenuation, has been investigated for over fifty years, beginning with the
work by Granato and Lücke [11]. A sketch of the basic mechanism describing the dependence of the anelastic strain
of dislocations pinned by core PDs on the applied stress is given in the inset of Fig. 3. The main hysteresis cycle is
given by the following equation

1
ean = S1 ൣ൫1+sgn(ɐ)sgn(ɐሶ )൯f(|ɐ|)+൫1-sgn(ɐ)sgn(ɐሶ )൯f(ɐmax)൧ɐ. (5)
2

The slope of the linear parts during unloading is determined by the fraction of dislocations that have broken away
from their pinning defects during loading. The simplifications implicit in the above model notwithstanding, the
resulting constitutive relation closely resembles the one which was proposed by Granato and Lücke (see, in particular,
Fig. 9 in [11]).
The modulus defect associated to this mechanism is given by

ɂ
οE/Eo ؆ - ቂ൫1+sgn(ɐ)sgn(ɐሶ )൯ ቀf(|ɐ|)+|ɐ|ɔ൫|ɐ|ห൯ቁ +൫1-sgn(ɐ)sgn(ɐሶ )൯f(ɐmax)൧. (6)
2

Figure 4 reports two examples illustrating the dependence of the relative modulus defect on the applied stress for
Vmax/' andassumingH  S1/So) = 1.
Note that a similar model of adhesion between faces of microcracks is able to account for only the maximum
during the loading phase, as one can realize comparing the constitutive relationships of a material containing
microcracks subjected to adhesion forces and eq. (6) [12]. Therefore, together with friction between cracks’ faces,
even adhesion may be excluded from those mechanisms capable of explaining DAET results in Berea sandstone.

-6 -4 -2 0 2 4 6
e 6
an ean 0
V'
V 4 -0.25
V V
cr max
2 -0.5

0 -0.75
-6 -4 -2 0 2 4 V6
-2 -1

-4 -1.25

-6 -1.5 'E/Eo

FIGURE 3. Anelastic strain versus stress for a distribution of FIGURE 4. Relative modulus defect associated to the
dislocations interacting with core PDs. The inset illustrates hysteresis cycle of Fig. 9 for two values of Vmax/' = 3, and 5.
the rheological model underlying the macroscopic strain-stress This figure has been produced assuming H = 1 in eq. (6).
relationship of a material with dislocations interacting with
core PDs.

Micro-Cracks with Finite Stiffness


A 1D simplified version of a model developed by Kachanov [13] to describe the effective properties of a solid
containing a distribution of crack holds that

1556
Ɏa2
e= ቂSo +N BH(ɐ)ቃ ɐ , (7)
V

where V is the representative volume of the micro-crack’s distribution, “a” is the crack radius, and B represents the
crack compliance. Equation (7) assume cracks that are circular and with the same radius. If the surfaces of the
microcracks are not perfectly conforming, upon compression they completely close only when the compressive
applied stress attains values larger than the material yield stress. To account for the partial closure of the cracks, eq.
(16) can be modified by allowing a crack to close only when V < Vcr < 0,

e=So ɐ൅S1 H(ɐ൅ɐcr )(ɐ൅ɐcr ) . (8)

Note that at V = 0, the microcracks are still open. The compliance S1 replaces the pre-factor N(Sa2)B/V. This equality
ensures that, once fully open, the COD of the microcracks is not affected by their surfaces’ topography. The linear
behavior of the strain for Vcr < V< 0 must be regarded only as a rough approximation to reality, aspects of which are
further discussed later. Finally, eq. (8) describes a purely elastic response, i.e., no hysteretic mechanism is
contemplated in it.
Averaging the extra-strain, 'e, over the distribution of Vcr (see eq. (2)) leads to the following constitutive relation

e = So ɐ൅S1 (|ɐ|൅ʹο)ൣH(ɐ)+H(-ɐ)eɐΤο ൧ . (9)

Figure 5 illustrates the behavior of the excess-strain in eq. (9). This result leads to a relative modulus defect, 'E/Eo,
given by

ɐ
(οEΤEo ) ؆-ɂ ቂH(ɐ)+H(-ɐ) ቀ1- ቁ eɐΤο ቃ, (10)
ο

whereH  S1/So). Note that averaging over the distribution of Vcr has introduced the qualitative exponential decay of
the excess-modulus caused by the progressive partial closure of interfaces between non-conforming, rough surfaces
for increasingly compressive stress values. Figure 6 shows the dependence of 'E/Eo on the normalized stress, which
has been evaluated using eq. (10) with H = 1. The decrease of the absolute values of the modulus defect for increasingly
compressive stress is the property of this model that it is proposed here to partially account for the overall negative
slope of the experimental cycles.

-6 -4 -2 0 2 4 6
10 0
'e / S1 V'
8 -0.25

6 -0.5

4 -0.75

2 -1

V' 'E / Eo
0 -1.25
-6 -4 -2 0 2 4 6

FIGURE 5. Averaged stress dependence of the extra-strain FIGURE 6. Relative modulus defect versus normalized
caused by a distribution of microcracks with finite stress for a distribution of microcracks with finite stiffness.
stiffness. The ratio 'e/S1 is given in units of Vmax. This figure has been produced assuming H = 1 in eq. (10).

1557
A Composite Model

The modulus defect, 'E/Eo, in Berea sandstone displays a hysteretic behavior with increasing complexity as the
amplitude of the excitation increases (Renaud et al. [2, 3] and Rivière et al. [4]). It is reasonable to assume that this
complex behavior is the results of the action of several mechanisms. In the previous section, two dissipative
mechanisms which may be described by constitutive relationships derived for the interaction between dislocations and
point defects and for distributions of micro-cracks with finite stiffness upon compression have been modeled. The
relative modulus defect recovered from the constitutive relationships which are linked to each of these three
mechanisms seems capable of accounting for at least one of the main features characterizing the experimental
observations. In fact,
i) the interaction between dislocations and glide PDs can be associated with the main hysteresis loops. The
constitutive relationship is an even (multivalued) function of V, and leads to a dependence of the intercept of
'E/Eo and of the nonlinear attenuation on 6 which resemble experimental evidence better than those based on
quadratic hysteresis. Furthermore, this mechanism can be shown to contribute to part of the average curvature
of 'E/Eo, the origin of which may otherwise be entirely attributed to a quadratic dependence of the modulus
defect on the strain (or stress) [4];
ii) hysteresis of 'E/Eo caused by the interaction between dislocations and core PDs may be associated to the local
maxima of 'E/Eo in proximity of the origin; finally,
iii) micro-crack distributions with finite stiffness, together with a quadratic contribution of classical origin, may
contribute to the variation of the average slope of 'E/Eo as the strength of the driving low-frequency wave
increases.
Since each mechanism (see eqs. (4), (6) and (10)) by itself modifies the linear behavior only by a small perturbation
of the material’s dynamics, their cumulative effect, in a first approximation, can be written as a weighted linear
combination of them all.

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2


0

V '

-0.2

-0.4
'E Eo)×10-2

-0.6

-0.8

-1

-1.2

FIGURE 7. Relative modulus defect for the composite model for three values of the stress amplitude,
Vmax / ' = 0.5, 1, and 1.5. The model’s parameters are given in Tables 1, 2, and 3.

1558
Figure 7 shows three examples of predictions by a model which includes all the above mechanisms, and a classical
contribution proportional to the second power of the stress. The parameters used to build the composite model are
given in Tables 1, 2, and 3, in which M1, M2, and M3 represent the models of sections 2.2, 2.3, and 2.4, respectively,
while “(stress)2” stands for the classical contribution proportional to V2. The symbol “W” refers to the weight given
to each model in the corresponding linear composition, and “'” denotes the width of the distribution of the
characteristic stress, Vcr, associated with each defect type. The normalization constant, 'o is the value of ' in the
model describing the distribution of Vcr for dislocations interacting with glide PDs (model M1). As one can infer from
earlier remarks, the weight W depends on the energy required to activate the associated mechanism. Disregarding
unphysical discontinuities, the cycles of Fig. 7 present similarities with data in Fig. 2 of [3] and Fig. 4 of [4] that can
hardly be fortuitous.
The present simulation has targeted experimental data that have been obtained for values of the pump’s amplitude
sufficiently high to activate all four mechanisms. The experimental data, however, make quite clear that hysteresis
associated to the interaction between dislocation and glide PDs appears much earlier than features linked to their
interaction with core PD. This observation is consistent with expectations based on the fact that core PDs offer greater
resistance to the dislocation’s motion than glide PDs. Similarly, in this type of rock, the effect of crack closure
becomes evident only at rather higher values of the pump wave’s amplitude, suggesting the presence of large pores
and/or fine debris preventing their closure. In short, each mechanism becomes active and produces observable effects
only after the excitation of the material has reached material-dependent threshold values.

TABLE 1. Parameters used to construct the composite model yielding the cycle with amplitude Vmax
/'o = 0.5 in Fig. 7. The symbol 'o represents the width of the distribution associated to the model M1.

M1 M2 M3 (stress)2
2
W[×10 ] 0.1125 0.01125 0.1875 0.225
''o 1 1 1.5

TABLE 2. Parameters used to construct the composite model yielding the cycle with amplitude Vmax
/'R = 1 in Fig. 7. The symbol 'o represents the width of the distribution associated to the model M1.

M1 M2 M3 (stress)2
2
W[×10 ] 0.1 0.06 0.45 0.075
''o 1 1 2

TABLE 3. Parameters used to construct the composite model yielding the cycle with amplitude Vmax
/'R = 1.5 in Fig. 7. The symbol 'o represents the width of the distribution associated to the model M1.

M1 M2 M3 (stress)2
2
W[×10 ] 0.0975 0.13 0.845 0.04875
''o 1 1 2

CONCLUSIONS
Constitutive relationships have been derived from rheological models of hysteretic and/or nonlinear elastic
interactions which describe important features of experimental findings in dry Berea sandstone obtained by means of
DAET. The rheological models may be associated to defects like dislocations interacting with point defects or sliding
of grain boundaries, and microcracks with finite stiffness. For each type of interaction, the associated modulus defect
has been derived from its constitutive relationship. A composite model, weighting the contribution of each
mechanism, has been introduced. Its predictions recover most of the main features of the experimental results

1559
qualitatively, and, in some case, theory and experimental evidence agree even in a quantitative sense. Whether the
interaction between dislocations and point defects is actually the mechanism responsible for the hysteretic behavior
of Berea sandstone needs to be further investigated. To this end, techniques alternative to DAET but with equal
sensitivity to variations of the elastic modulus should be employed.

REFERENCES
1. G. Gremaud, “Dislocations - point defects interactions”, Materials Science Forum 366-368, 178-246 (2001).
2. G. Renaud, P.-Y. Le Bas, and P.A. Johnson, “Revealing highly complex elastic nonlinear (anelastic) behavior of
Earth materials applying a new probe: Dynamic acoustoelastic testing”, J. Geoph. Res. 117, B06202 (2012).
3. G. Renaud, J. Rivière, P.-Y. Le Bas, and P.A. Johnson, “Hysteretic nonlinear elasticity of Berea sandstone at
low-vibrational strain revealed by dynamic acousto-elastic testing”, Geoph. Res. Lett. 40, 715–719 (2013).
4. J. Rivière, G. Renaud, R.A. Guyer, and P.A. Johnson, “Pump and probe waves in dynamic acousto-elasticity:
Comprehensive description and comparison with nonlinear elastic theories” J. Appl. Phys. 114, 054905 (2013).
5. L.A. Ostrovsky, and P.A. Johnson, “Dynamic nonlinear elasticity in geomaterials”, Rivista del Nuovo Cimento
24, 1-46 (2001).
6. O.O. Vakhnenko, V.O. Vakhnenko, T.J. Shanland, and J.A. Tencate, “Strain-induced kinetics of intergrain
defects as the mechanism of slow dynamics in the nonlinear resonant response of humid sandstone bars”,
Physical Review E, 70, 015602 (2004).
7. V. Aleshin, and K. Van Den Abeele, “Microcontact-based theory for acoustics in microdamaged materials”, J.
Phys. Mech. Solids 55, 366-390 (2007).
8. V. Lyakhovsky, Y. Hamiel, J.-P. Ampuero, and Y. Ben-Zion, “Non-linear damage rheology and wave resonance
in rocks”, Geophys. J. Int. 178, 910-920 (2009).
9. C. Pecorari, “Nonlinear interaction of plane ultrasonic waves with an interface between rough surfaces in
contact”, J. Acoust. Soc. Am. 113, 3065-3072 (2003).
10. C. Pecorari and D. A. Mendelsohn, “Modeling the Nonlinear Hysteretic Response of Distributed Damage in a
1D Resonance Experiment”, in Review of Progress in Quantitative Nondestructive Evaluation, ed. D. E.
Chimenti, L. J. Bond, and D. O. Thompson, (American Institute of Physics 1581, Melville, NY) 33, 703-710
(2014).
11. A.V. Granato, and K. Lücke, “Theory of mechanical damping due to dislocations”, J. Applied Physics 27, 583-
593 (1956).
12. C. Pecorari, and D.A. Mendelsohn, “Forced nonlinear vibrations of a one-dimensional bar with arbitrary
distributions of hysteretic damage”, Journal of Nondestructive Evaluation 33, 239-251 (2014).
13. M. Kachanov, “Effective elastic properties of cracked solids: critical review of some basic concepts”, Appl.
Mech. Rev. 45, 304-335 (1992).

1560

You might also like