Professional Documents
Culture Documents
Technical Communication
a r t i c l e i n f o a b s t r a c t
Article history: The Mohr–Coulomb, Tresca and Von Mises criteria are classical failure criteria, widely accepted and used
Received 17 January 2008 for various materials. To take into account the influence of intermediate principal stress on the strength
Received in revised form 8 September 2008 of soil, Bishop, Lade–Duncan and Matsuoka proposed criteria in terms of three principal stresses or three
Accepted 10 September 2008
stress invariants. This note describes an expression with two parameters for modeling the shape of the
Available online 23 October 2008
failure surface in the octahedral plane. By studying the roles of these two parameters and comparing
the new criterion with the aforementioned criteria, the advantage and flexibility of the proposed function
Keywords:
is explored. Application of the function is demonstrated by fitting the new surface to experimental data
Failure surface
Convexity
for various soils.
Friction angle Ó 2008 Elsevier Ltd. All rights reserved.
Stress ratio
Lode angle
0266-352X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2008.09.001
518 D. Su et al. / Computers and Geotechnics 36 (2009) 517–524
Nomenclature
u effective internal friction angle Rc, Re failure stress ratio in triaxial compression and in triaxial
r1, r2, r3 effective principal stresses extension
b a measurement of intermediate principal stress r normalized stress ratio
J1, J2, J3 stress invariants h the Lode angle
p effective mean normal stress a, m parameters in the shape function
sij deviatoric stress tensor
J2D, J3D second and third invariants of the deviatoric stress
tensor
0.2 It is worth mentioning that for the case m = 1, the basic function re-
0 duces to
2a
-0.2 r0 ðhÞ ¼ ð11Þ
1 þ a þ ð1 aÞ sin 3h
-0.4
which is equivalent with the equation proposed by Argyris et al. [2].
-0.6 For the case m = 4, the basic function is
-0.8 1=4
2a4
r0 ðhÞ ¼ ð12Þ
-1 1 þ a4 þ ð1 a4 Þ sin 3h
-1.2 which is equivalent with the expression proposed by Sheng et al.
-1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2
[18] and it is used in a generalized version of the modified Cam-clay
Fig. 1. Classical failure criteria. model. It can be demonstrated that the form in Eq. (11) is not valid
D. Su et al. / Computers and Geotechnics 36 (2009) 517–524 519
0.6 and
1=m " 2 #
2am cos 3h
0.4 r¼ 1 ¼ a at h ¼ 30
1 þ am þ ð1 am Þ sin 3h 6m3=2
0.2
-0.6
" 2 #
3ð1 am Þ cos 3h cos 3h
1
-0.8 2mam 6m3=2
1þ1=m
-1 2am
ð13Þ
-1.2 1 þ am þ ð1 am Þ sin 3h
-1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2
for values of a less than 7/9, due to the concave part around h = 30°, 1
while the form in Eq. (10) with m = 1 can ensure the convexity 0.8
when a is not less than 0.716 (Fig. 2). In the following section, the
restriction imposed upon the values of m, a in Eq. (10) due to con- 0.6
0.2
3. Discussion
0
According to the aforementioned definitions, an expression for -0.2
failure surfaces should have the following features:
-0.4
-1.2
Conditions (1) and (2) are basically required in defining failure -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2
surfaces in terms of the normalized stress ratio in Eq. (4). Condition
(3) reflects the symmetric characteristics of failure surfaces along Fig. 4. Failure surfaces with m = 4.2.
the principal stress axes on the octahedral plane (yield surfaces
for anisotropic material show different characteristics, as pre-
sented by Abelev and Lade [1], and are not discussed here), as well 1.2
as a unique normal direction on the points where yield surfaces
intercept the principal axes (the Mohr–Coulomb criterion shows 1
a distinct feature that there exist non-unique normal directions
0.8
at the intersection points of failure surfaces and the axes). Condi-
tion (4) is to make sure that the law of thermodynamic is not vio- 0.6 m =1, 3, 5, 7, 9
lated when the surface is used as a yield surface.
0.4
0.2
1
0
0.9
-0.2
0.8
Convex -0.4
α
0.7 -0.6
0.6 -0.8
Concave
-1
0.5
1 2 3 4 5 6 7 8 9 10 -1.2
m -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2
Fig. 3. Values of m and a that ensure convexity of failure surfaces. Fig. 5. Failure surfaces with a = 0.7.
520 D. Su et al. / Computers and Geotechnics 36 (2009) 517–524
It can be readily seen from Eq. (13) that r0 (h) = 0 at both h = 30° The value of a that satisfies Eq. (14) can be found numerically when
and h = 30°. m is specified. The results for the value of m between 1 and 10 are
To ensure the convexity of the shape function, the following shown in Fig. 3. It is seen that the minimum allowable value of a is
equation: 0.716 when m = 1, but the value varies with m increasing. The lower
limit of the value is 0.611 occurring at m = 4.2. As shown in Fig. 4,
ðrðhÞÞ2 þ 2ðr 0 ðhÞÞ2 rðhÞ r 00 ðhÞ 0 ð14Þ failure surfaces with a > 0.611 are convex while those with
a < 0.611 are concave when m = 4.2. For a specified a, the degree
should be satisfied [6], where the second derivative is
of convexity of curves depends on the value of m. For example,
1=m for a being 0.7, the curves with m = 3, 5, 7 are convex, but those
cos 6h 2am
r 00 ðhÞ ¼ with m = 1, 9 are concave, as shown in Fig. 5. The critical m values
2m3 1 þ am þ ð1 am Þ sin 3h
can be obtained from Fig. 3. Compared with the criteria proposed
ð1 am Þ sin 3h½36m3 5ðcos 3hÞ2 by Argyris et al. [2], Wang et al. [21], and Sheng et al. [18], in which
þ
8m4 am the shape of failure surface on the octahedral plane is uniquely
1þ1=m
2am determined by the value of a, the proposed criterion can cover
1 þ am þ ð1 am Þ sin 3h much wider range of soil behaviors with the use of the parameter m.
ðm þ 1Þð1 am Þ2 ðcos 3hÞ2 ½36m3 ðcos 3hÞ2
þ 4. Comparison with the SMP and Lade–Duncan criteria
16m5 a2m
m
2þ1=m
2a
ð15Þ A distinctive feature of the SMP criterion is that, like Mohr–
1 þ am þ ð1 am Þ sin 3h
Coulomb model, it predicts the same friction angles in triaxial
65
Lade-Duncan
60
55
SMP
50
Friction angle (º)
Lade-Duncan
45
SMP
40
Lade-Duncan
35
30 SMP
25
20
-30 -20 -10 0 10 20 30
Lode angle (º)
Fig. 6. Relationships between the Lode angle and friction angle in the Lade–Duncan and SMP criteria.
70 30
Lode angle at maximum friction angle (º)
60 20
Maximum friction angle (º)
Lade-Duncan
50 10
SMP
40 0
Lade-Duncan
SMP
30 -10
20 -20
Maximum friction angle
Lode angle at maximum friction angle
10 -30
25 30 35 40 45 50
Friction angle at triaxial compression (º)
Fig. 7. Variations of maximum friction angle (and associated Lode angle) with friction angle at triaxial compression in the Lade–Duncan and SMP criteria.
D. Su et al. / Computers and Geotechnics 36 (2009) 517–524 521
compression (h = 30°) and in triaxial extension (h = 30°), while dicted friction angle versus the Lode angle obtained from the
the Lade–Duncan criterion always predicts larger values in triaxial SMP and Lade–Duncan criteria, with the triaxial compression fric-
extension than in triaxial compression. Fig. 6 presents the pre- tion angle being 30°, 40°, and 50°, respectively.
60
55
Friction angle (º)
50
45
40
α =0.8, 0.709, 0.7, 0.647, 0.625
35
30
-30 -20 -10 0 10 20 30
Lode angle (°)
Fig. 8. Influence of a on the relationship between the Lode angle and the friction angle when m = 5.
60
55
Friction angle (º)
50
45
m =3, 5, 7
40
35
30
-30 -20 -10 0 10 20 30
Lode angle (º)
Fig. 9. Influence of m on the relationship between the Lode angle and the friction angle when a = 0.7.
90
80
Maximum friction angle (º)
70
60
50
40
α =0.8, 0.7, 0.625
30
20
25 30 35 40 45 50
Friction angle at triaxial compression (º)
Fig. 10. Influence of a on the relationship between the maximum friction angle and the friction angle at triaxial compression when m = 5.
522 D. Su et al. / Computers and Geotechnics 36 (2009) 517–524
The experimental results by Kirkpatrick [7], Sutherland and h = 30°. On the other hand, the results by Green and Bishop [5],
Mesdary [20], Ramamurthy and Rawat [17] showed that the inter- Proctor and Barden [16], Lade and Duncan [8] showed that the
nal friction angles at h = 30° were approximately equal to that at internal friction angle at h = 30°, i.e. triaxial extension, is larger
90
80
Maximum friction angle (º)
70
60
50
40
m =3, 5, 7
30
20
25 30 35 40 45 50
Friction angle at triaxial compression (º)
Fig. 11. Influence of m on the relationship between the maximum friction angle and the friction angle at triaxial compression when a = 0.7.
0 0 0 0
-0.6
σ2 σ3 -0.6
σ2 σ3 -0.6
σ2 σ3 -0.6
σ2 σ3
-0.8 -0.8 -0.8 -0.8
-1 -1 -1 -1
0 0 0 0
-0.6 σ2 σ3 -0.6
σ2 σ3 -0.6
σ2 σ3 -0.6
σ2 σ3
-0.8 -0.8 -0.8 -0.8
-1 -1 -1 -1
0 0 0 0
-1 -1 -1 -1
Fig. 12. Failure surfaces for various test data from triaxial, true triaxial and thick cylinder tests.
D. Su et al. / Computers and Geotechnics 36 (2009) 517–524 523
than that at triaxial compression (h = 30°). Hence, the SMP and 5. Applicability of the shape function
Lade–Duncan criteria might be right, but neither of them can de-
scribe the whole picture. Furthermore, it can be seen from Fig. 6 With conventional triaxial apparatus, only triaxial compression
that the shape of failure curves from these two criteria is uniquely (r1 > r2 = r3) and triaxial extension tests (r1 = r2 > r3) can be per-
determined by the friction angle in triaxial compression (ucom), formed, and limited information regarding the shearing strength
implying that the influence of the Lode angle (i.e. the intermediate may be provided. To investigate the influence of the intermediate
principle stress) on the internal friction angle is only determined principal stress on the shearing strength, true triaxial and hollow
by ucom. A unique relationship between the maximum friction an- cylinder devices have been utilized by a number of researchers,
gle (umax) and ucom was also assumed in these two criteria, as such as Proctor and Barden [16], Ramamurthy and Rawat [17],
shown in Fig. 7. The umax, given by the Lade–Duncan criterion, is and Matsuoka and Nakai [14]. In true triaxial apparatus, the three
always larger than that given by the SMP criterion. However, the principal stresses (r1, r2, r3) can be controlled independently,
Lode angle at which umax occurs is a constant, being 0° according while the vertical load, the torsional load, the internal pressure
to the Lade–Duncan criterion, but that change linearly with ucom and the external pressure can be applied independently in hollow
according to the SMP criterion, which is always smaller than 0°. cylinder devices. Using a combination of these two devices, a vari-
Compared with the SMP and Lade–Duncan criteria, the pro- ety of stress paths can be executed. Therefore, the proposed failure
posed criterion offers more flexibility in serving as the failure cri- criterion is fitted to the data obtained from various experiments
terion as the two parameters m and a in the function provide the performed on true triaxial and hollow cylinder devices, as shown
ability to better control the shape of the failure surface. Figs. 8, 9 in Fig. 12(a–l). The corresponding parameters m and a of the curves
give examples of how the friction angles varies for ucom = 40°. To are listed in Table 1. It can be seen that although ucom ranges from
illustrate the effect of a on the relationship between the Lode angle 31.9° to 48.5° and uext ranges from 31.9° to 57.2° in the data, all
and the friction angle, m is fixed at 5.0 in Fig. 8. It can be seen that cases (ucom > uext, ucom = uext and ucom < uext) can be reasonably
the criterion can take into account all cases with uext > ucom, represented by selecting proper values of m and a.
uext = ucom and uext < ucom by adjusting the value of a. The curve
given by the SMP criterion with ucom = 40° in Fig. 6 is close to 6. Conclusions
the curve with a = 0.647, while that given by the Lade–Duncan
criterion is close to the curve with a = 0.709. In Fig. 9, a is a con- An expression for the shape of failure surfaces in the octahedral
stant, being 0.7, which means the ratio of stress ratio at h = 30° plane is proposed, which is a function of stress invariants and two
to that at h = 30° is unchanged. But, with varying m, different parameters: the failure stress ratio in triaxial extension over that in
curves with different umax can be obtained. From Figs. 10 and 11, triaxial compression a, and a new parameter m controlling the con-
it can be seen that umax increases with the value of a increasing, vexity of the surface. This function can take into account the influ-
but decreases with the value of m increasing, being more sensitive ence of the intermediate principal stress on strength of materials
to the former. Although situations with umax = 90°, such as a = 0.8 and offers greater flexibility than the classical criteria in describing
in Fig. 10, are not likely to occur, the proposed criterion can take the strength characteristics of a material. It is believed that the cri-
into consideration of such extreme cases. terion has potential applications in developing constitutive rela-
tionships for geomaterials and other materials.
Table 1
Parameters of failure surfaces for various test data Acknowledgements
[13] Matsuoka H, Nakai T. Stress–strain relationship of soil based on the ‘SMP’. In: [19] Shibata T, Karube D. Influence of the variation of the intermediate principal
Proceedings of specialty session 9, 9th international conference on soil stress on mechanical properties of normally consolidated clay. In: Proceedings
mechanics and foundation engineering; 1977: p. 153–62. of 6th international conference on soil mechanics and foundation engineering,
[14] Matsuoka H, Nakai T. A new failure criterion for soils in three-dimensional vol. 1; 1965: p. 359–63.
stresses. In: IUTAM conference on deformation and failure of granular [20] Sutherland HB, Mesdary MS. The influence of the intermediate principal stress
materials, Delft, vol. 1; 1982: p. 253–63. on the strength of sand. In: Proceedings of 7th international conference on soil
[15] Motara G. A new yield and failure criterion for geomaterials. Géotechnique mechanics and foundation engineering, Mexico City, vol. 1; 1969: p. 391–99.
2008;58(2):125–32. [21] Wang ZL, Dafalias YF, Shen CK. Bounding surface hypoplasticity model for
[16] Proctor DC, Barden L. A note on the drained strength of sand under generalized sand. J Eng Mech, ASCE 1990;116(5):983–1001.
strain conditions. Géotechnique 1969;19(1):144–9. [22] Yamada Y, Ishihara K. Anisotropic deformation characteristics of sand under
[17] Ramamurthy T, Rawat PC. Shear strength of sand under general stress system. three-dimensional stress conditions. Soils and Foundations 1979;19(2):78–94.
Proc 8th Int Conf Soil Mech Found Eng Moscow 1973;1(2):339–42. [23] Yu M. Advances in strength theories for materials under complex stress state
[18] Sheng D, Sloan SW, Yu HS. Aspects of finite element implementation of critical in the 20th Century. Applied Mechanics Reviews, ASME 2002;55(3):169–218.
state models. Comput Mech 2000;26(2):185–96.