Вы находитесь на странице: 1из 8

Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 191 (2017) 488 – 495

Symposium of the International Society for Rock Mechanics

Sensitivity Analysis of the Micro-Parameters Used in a PFC


Analysis Towards the Mechanical Properties of Rocks
U. Castro-Filgueiraa,*, L.R. Alejanoa, J. Arzúab, D. Mas Ivarsc
a
Natural Resources & Environmental Engineeting Department, University of Vigo, Spain
b
Department of Metallurgical and Mining Engineering, Universidad Católica del Norte, Chile
c
Swedish Nuclear Fuel and Waste Mangement Co, SKB, Stockholm, Sweden

Abstract

Reproduction on complex behavior of rocks in actual case studies is undoubtedly the main aim of numerical modelling applied to
rock mechanics. However, to achieve a reliably realistic simulation process a good knowledge of rock behavior is needed
together with calibration or matching processes of actual and controlled tests on rock. Within this framework, in a previous
approach, the authors calibrated PFC (Particle Flow Code) micromechanical parameters to reproduce in a reasonably accurate
way the complete stress-strain response of granite samples under triaxial conditions in the lab. In this previous study, it was
found that the number of different micro-parameters needed to define a PFC model was large, and that the influence of some of
these parameters on rock response was not clear. In this new study, a sensitivity program is carried out in order to assess
the influence of micro-mechanic parameters on the macroscopic response of the numerical simulation of rock samples.
The sensitivity approach carried out shows that there is a direct connection between some parameters, for example, the bond
cohesion and the bond tensile strength influences the peak strength and tensile strength of the rock, respectively. However, it is
not clear how these parameters may influence other geomechanical macro properties, neither the exact influence of other
parameters as the friction coefficient or the stiffness ratio. The present study addresses some of these issues, and it sheds light on
which are most relevant parameters and how they control the main macro-properties of the simulated rocks.
© 2017
© 2017TheTheAuthors.
Authors. Published
Published by Elsevier
by Elsevier Ltd. Ltd.
This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of EUROCK 2017.
Peer-review under responsibility of the organizing committee of EUROCK 2017
Keywords: PFC; flat-joint; microproperties

* Corresponding author. Tel.: +34-986-812-374.


E-mail address: ucastro@uvigo.es

1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of EUROCK 2017
doi:10.1016/j.proeng.2017.05.208
U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495 489

1. Introduction

During last years, numerical modelling has increased its use in rock engineering approaches in a significant
manner, becoming an important tool both in research and project design. In the field of rock mechanics research,
this tool has been used both to model real situations and to reproduce different tests carried out at laboratories [1].
This reproduction of the laboratory tests focuses to expand our knowledge on rock behavior as to help us to give an
answer to why the rock behaves as it does.
In the Rock Mechanics laboratory of the University of Vigo, the stress-strain behavior of different rocks has been
studied, obtaining the different parameters that define the pre- and post-peak behavior, with special attention to
the last one, due to the scarce information usually available in this field [2, 3]. With the aim of simulating rock
compressive strength tests, it was decided to carry out simulations in order to better understand this behavior and to
learn how to estimate the micro-mechanic parameters capable to reproduce the actual macroscopic behavior of rock.
To do these simulations, the code PFC (Particle Flow Code) of the company Itasca Consulting Group was chosen.
When trying to model this behavior, it was found that the number of different micro-parameters needed to define
a PFC model is not only large, but also that influence of these parameters in behavior is not completely clear. This
paper presents a study of the influence of different micro-properties that define the flat-jointed material of
the Particle Flow Code.

2. Calibration approaches when modelling with PFC

The number of micro-properties, which define the contact models and are required to calibrate the models, is
quite large. There are different calibration procedures as suggested by Itasca [4] (Fig. 1) or the suggested by Wu and
Xu [5] (Fig. 2). In both procedures, authors calibrate same properties, but the order of calibration of the parameters
are quite different. Moreover, there are different genesis procedures of the model. On one hand, the procedure
suggested by Potyondy and Cundall [6], where the packing arrangement and the properties assignment occur at
the same time. On the other hand, the suggested by Shöpfer et al. [7], where first the packing arrangement is created
with MatLab and then imported to PFC in order to install the model properties.

Fig. 1. Calibration procedure defined by Itasca [4].

However, for the actual version of PFC (PFC 5.0) there are no studies about how the different parameters that
define the models influence on the response of the simulations. Although Wu and Xu [5] show in their paper a little
study of the influence of some parameters (average coordination number, crack density, bond cohesion to tensile
strength ratio local friction angle and residual friction angle), they study their influence exclusively in peak strength
parameters such as the UCS / ıt ratio, Hoek-Brown strength parameter mi, and internal friction angle, working with
the previous version of PFC (PFC 4.0). So, the present study shows the influence of the main micro-properties of
the flat-joint model that have a direct influence on the deformability and the peak and residual strength of the rock.
There are a number of other parameters that may also influence the genesis procedure of the model, but their
influence is considered out of the scope of this study.
490 U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495

Fig. 2. Calibration procedure defined by Wu and Xu [5].

3. Numerical simulation of stress-strain behavior of intact granite with PFC

In a previous study [8], the authors carried out a calibration numerical process with code PFC and following
the procedure suggested in [4] to be able to model the complete stress-strain behavior of intact granite samples.
These complete stress-strain curves were fitted to a good number of laboratory tests on a granitic rock previously
performed in [2] by two of the authors of this paper. The authors tried first a parallel bond material model [8], to
find out that it was not possible to simulate the actual frictional characteristics of the rock. This was why a flat-joint
model [8] was considered more convenient to achieve a better reproduction of rock behavior.
Based on this flat joint model, a calibration procedure in line with that suggested by Itasca (Fig. 1) was followed
in order to fit the micro-mechanical properties needed as the input parameters. The final micro-mechanical input
properties used in this initial approach are those presented in Table 1. By using these parameters, it was possible to
reproduce in reasonably approximate manner the actual response of granite samples in terms of the elastic, peak and
post-peak parameters. Fig. 3 shows numerical and laboratory stress strain curves of triaxial tests of granite samples
submitted to confinement stress of 2 (Fig 3a) and 10 MPa (Fig. 3b), respectively.

Table 1. Micro-properties for Blanco Mera granite.


Microproperty Description Value
Dmax/Dmin 1.66
Nr Nº elements in radial direction 1
NĮ Nº elements in circumferential direction 3
g0 Installation gap [mm] 0.5
‫׋‬B Bonded fraction 1.0
‫׋‬G Gapped fraction 0.0
ȡ Density [kg/m3] 2600
E* Effective modulus [GPa] 84
k* Stiffness ratio 1.7
ȝ Friction coefficient 0.5
Ȝ Radius-multiplier 1.0
E* Effective modulus [GPa] 56
k* Stiffness ratio 1.7
ıc Contact tensile strength [MPa] 4
c Contact cohesion [MPa] 255
‫׋‬ Friction angle [°] 0.5
U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495 491

As it can be observed, whereas the match in the pre-failure phase of deformation and up to peak is quite accurate,
the matching level is not that high in the last stages of the post-failure phase. In order to improve the accuracy of
the matching process it was deemed interesting to carry out a sensitivity analysis, with the aim of clear the influence
of input micro-properties in output results in terms of the most relevant geomechanical parameters derived from
the models, including Young Elastic Moduli, peak and residual strength and dilatancy.

Fig. 3. Stress-strain curves for 2 and 10 MPa confinement stress compressive tests as derived from laboratory (solid line) and as obtained with
PFC3D (flat-joint approach) (dotted line) together with main geomechanical parameters to be analayzed.

4. Flat-jointed material

The flat-jointed material for hard rock was defined by Potyondy [9], and it consists of an arrangement of
particles, which in our case are balls, connected each other in their contact points by flat-joint contacts (Fig. 4).
The grains of the flat-joint material are faced, depicted as a spherical core with a number of skirted faces. The faced
grains are created when the flat-joint contact model is installed at their contact points. The flat-joint contact model
simulates the behavior of an interface between two particles with locally flat notional surfaces. The interface
coincides with a middle surface that remains centered on the contact plane, its mechanical behavior is either
frictional or bonded, and may vary along the interface.

Fig. 4. Flat-joint contact (left) and flat-joint material (right) [4].


492 U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495

The flat-joint model provides the macroscopic behavior of a finite-size, linear elastic and either bonded or
frictional interface that may sustain partial damage (Fig. 5). This interface is discretized into elements, each of
which is either bonded or unbonded. The behavior of a bonded element is linear elastic until the strength limit is
exceeded and the bond breaks, making the element unbonded; the behavior of an unbonded element is linear elastic
and frictional, with slip accommodated by imposing a Coulomb limit on the shear.

Fig. 5. Behavior and rheological components of the flat-joint model [4].

As we can see, the microproperties that controls the flat-join model are: the contact tensile strength (ıc),
the contact cohesion (c), the friction angle (‫)׋‬, the friction coefficient (ȝ), shear stiffness (ks) and normal stiffness
(kn). When we define the flat-joint model, the shear and normal stiffnesses are not defined directly. In its stead, the
effective modulus (E*) and the normal-to-shear stiffness ratio (k*) are used, and they are related as follows:

k : E ; k : kn *
(1)
n L s k

Where L is the addition of the radios of the balls in contact.

5. Influence of micro-properties

Although there are several parameters that define a flat-jointed material, we have decided to study initially those
which controls the deformability and the strength of the simulated rock, it means, the parameters which controls
the behaviors of the flat-joint contacts and the balls. These parameters are (as shown in Table1): the contact
cohesion, the contact tensile strength, the effective moduli of the balls and contacts, the friction angle, the friction
coefficient and the stiffness ratios of the balls and contacts. For the study of each micro-property, compressive
strength tests with 2 MPa of confinement and tensile tests have been carried out. For each changed value, ten
different packing arrangements were created, in order to take into account the variability of the macro-properties as
happen in real tests. The studied macro-properties of the rock are: Young’s modulus; Poisson’s ratio; peak, residual
and tensile strengths, and the dilation angle associated to the peak and to the 75% of the difference between and
residual strengths. All the results presented below, and represented in Fig.6, are standardized with respect to
the results obtained in the calibration of the granite. It is necessary to take into account that graphs in Fig. 6 are used
different have different scales on y-axis in order to better represent the influence of each micro-properties.
In Table 2, we present the maximum percentage variation of the macro-properties corresponding to each micro-
property, calculated for the range considered in each micro-property. The information provided in rows of Table 2
together with the corresponding graph of Fig. 6 helps to understand the actual variation of each macro-property,
since different y-axis scales have been used in Fig.6 for illustrative purposes.
U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495 493

Fig. 6. Influence of the different micro-properties over the macro-properties of the rock.

Table 2. Maximum percentage variation of macro-properties (columns) for each micro-property (rows) variated.
Young’s Modulus Poisson’s ratio Peak strength Residual strength Tensile strength Dilatancy
Contact cohesion -8% 70% 80% 80% 0% 14%
Contact tensile strength 3% -15% 5% 6% >100% -3%
Effective modulus 42% 1% 1% 1% -2% -4%
Friction angle -1% 5% 6% 6% 0% 1%
Friction coefficient -1% 11% 13% 12% 0% 15%
Stiffness ratio -46% >100% -29% -29% -22% 22%
494 U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495

5.1. Contact cohesion

Contact cohesion has been varied in a range from 150 MPa to 355 MPa, including the results obtained for
the calibration parameters. As expected (Fig. 6a and Table 2), the contact cohesion has a direct influence over
the peak and residual strengths. The contact cohesion has also a direct influence on the Poisson’s ratio and dilation
angle, and an indirect relation with the Young’s modulus. However, this parameter has no impact on the tensile
strength of the rock, the results we obtain in the simulations were exactly the same for each contact cohesion value.

5.2. Contact tensile strength

As for the contact cohesion, compressive strength and tensile tests were carried out with different values of
the contact tensile strength, from 2 to 6 MPa. In this case, the contact tensile strength has direct impact on the tensile
strength of the rock (Fig. 6b and Table 2), and also has a slight influence on the Young’s modulus and the peak and
residual strengths. However it has an indirect influence on the Poisson’s ratio and a little influence over
the dilatancy.

5.3. Effective moduli

In the flat-joint material, we have the effective moduli of the balls and the contacts. These parameters were
usually stablished identical, however due to authors’ experience, the ball effective modulus is 1.5 times the contact
effective modulus. In this case, we vary the ball effective modulus from 70 to 100 GPa, therefore the contact
effective modulus with the previously explained relation. From the results in Fig. 6c and Table 2, we can observe
that the effective moduli seems to have only influence on the Young’s modulus of the rock, not affecting sensitively
to other macroproperties.

5.4. Friction angle

The friction angle was varied from 0.0° to 10.0°, and the results obtained from the triaxial and tensile tests are
shown in Fig. 6d and Table 2. As we can observe from the results, this parameter has not a higher influence over
the macroproperties. The friction angle slightly increase the Poisson’s ratio, the dilation angle and the peak and
residual strengths, and slightly decrease the Young’s modulus. As with the contact cohesion, the friction angle has
no influence on the tensile strength, obtaining identical results for all friction angle values

5.5. Friction coefficient

The friction coefficient has been varied from 0.0 to 1.0, and the results obtained from the different tests carried
out are represented in Fig. 6e and Table 2. The friction coefficient only has a remarkable influence on the peak and
residual strength, the dilatancy and on the Poisson’s ratio. The Young’s modulus slightly decrease and the tensile
strength in not affecting not at all, as with the contact cohesion and the friction angle, we obtain the exact results in
all tests.

5.6. Stiffness ratio

As in the effective moduli, the flat-jointed material has a ball stiffness ratio and a contact stiffness ratio. In this
case, we have stablished equal these parameters and varied them from 1.0 to 2.5. The results of this analysis are
shown in Fig. 6f and Table 2. From these results, we can observe that these parameter has a high influence on all
the macroproperties of the rock. The Poisson’s ratio is the most influenced parameter. This property and dilatancy
increase their values when stiffness ratio increases. The remaining macroproperties decrease when the stiffness ratio
increases.
U. Castro-Filgueira et al. / Procedia Engineering 191 (2017) 488 – 495 495

6. Conclusions

In this paper, authors have presented the results of a sensitivity analysis of the main microproperties that have
a direct impact on the mechanical behavior of rocks. From the reached results for the contact cohesion, we can
observe that has a large influence on the peak and residual strengths, as expected, and on Poisson’s ratio, and an
indirect relation with the Young’s modulus. However, this microproperty has no influence over the tensile strength
of the rock. The contact tensile strength has a direct influence over the tensile strength of the rock, as expected, and
also an indirect influence over the Poisson’s ratio. We can observe that this parameter has also a little influence over
the Young’s modulus and the peak and residual strengths. The effective moduli of the balls and contacts only affects
the Young’s modulus of the rock, as the elasticity of the balls and contacts increase, the elasticity of the response
increases. However, this parameter do not affects sensitively the other macroproperties. In the case of the friction
angle and the friction coefficient, both parameters have little influence. As they increase, peak and residual strengths
increase, Poisson’s ratio is hardly increased, the Young’s modulus has not a clear influence, and the tensile strength
is not influenced at all. Finally, the stiffness ratio is the only parameter that has a clear influence on all
the macroproperties of the rock. This parameter, when increases, only the Poisson’s ratio increases, the other
macroproperties decrease.
It is interesting to note, that around 30 variations of the parameters were carried out, which is equivalent to more
than 30 days of PFC computer work on the models. Besides that, all the obtained results had to be computed and
analyzed afterwards.
Particular relevant, in what concerns the accuracy of simulation results in the post-peak stage, is the variation of
dilatancy with input parameters. In particular, it is observed that dilatancy increases with friction coefficient and
stiffness ratio, and hardly varies with the other input parameters analyzed. According to this results, slightly
diminishing these input parameters would produce less dilatancy line with actual observations in test results. Future
work involves trying to better calibrate input parameters with the aim of better represent dilatant behavior.

Acknowledgements

The authors acknowledge Itasca Consulting Group for accepting the first author in their Itasca Education
Partnership Program, which provides her with the PFC program. The authors also thank the Spanish Ministry of
Economy and Competitiveness for financial support of the laboratory study, awarded under Contract Reference No.
BIA2014-53368P. This contract is partially financed by means of ERDF funds of the EU.

References

[1] D. Mas Ivars, M.E. Pierce, C. Darcel, J. Reyes-Montes, D.O. Potyondy, R.P. Young, P.A. Cundall, The synthetic rock mass approach for
jointed rock mass modelling, Int. J. Rock Mech. & Min. Sci. 48 (2011) 219–244.
[2] J. Arzúa, L.R. Alejano, Dilation in granite during servo-controlled triaxial strength tests, Int. J. Rock Mech. & Min. Sci. 61 (2013) 43–56.
[3] J. Arzúa, L.R. Alejano, G. Walton, Strength and dilation of jointed granite specimens in servo-controlled triaxial tests. Int. J. Rock Mech.
& Min. Sci. 69 (2014) 93–104.
[4] Itasca, PFC (Particle Flow Code) Version 5.0. Itasca Consulting Group Inc. Minneapolis, United States, 2014
[5] S. Wu, X. Xu, A Study of Three Intrinsic Problems of the Classic Discrete Element Method Using Flat-Joint Model, Rock Mech. & Rock Eng.
49 (2016) 1813–1830.
[6] D.O. Potyondy, P.A. Cundall, A bonded-particle model for rock, Int. J. Rock. Mech. & Mining Sci. 41 (2004) 1329–1364.
[7] M.P.J. Schöpfer, S. Abe, C. Childs, J.J. Walsh, The impact of porosity and crack density on the elasticity, strength and friction of cohesive
granular materials: Insights from DEM modeling, In. J. Rock. Mech. & Min. Sci. 46 (2009) 250–261.
[8] U. Castro-Filgueira, L.R. Alejano, J. Arzúa, D. Mas Ivars, Numerical simulation of the stress-strain behavior of intact granite specimens with
Particle Flow Code, Proc. EUROCK 2016 / 2016 ISRM International Symposium, Cappadocia, Turkey, 2016.
[9] D.O. Potyondy, A Flat-Jointed Bonded-Particle Material for Hard Rock, 46th US Rock Mech/Geomech Symposium 2012, 3 (2012)
1510–1519.

Вам также может понравиться