Você está na página 1de 7

Martin, C. M.

(2011)

Ge otechnique Letters 1, 2329, http://dx.doi.org/10.1680/geolett.11.00018

The use of adaptive finite-element limit analysis to reveal slip-line fields


C. M. MARTIN* The numerical method known as finite-element limit analysis (FELA) is generally employed as a tool for obtaining lower and upper bounds on the exact collapse load of a perfectly plastic structure or continuum. Most applications of FELA in geotechnical engineering have focused on plane strain problems involving the classical Tresca and MohrCoulomb yield criteria, and considerable computational effort has been expended on the calculation of lower- and upper-bound solutions for particular problems. This paper discusses and demonstrates an alternative use of FELA as a tool for ascertaining slip-line fields for plane strain problems. A simple but effective strategy for adaptive mesh refinement is a key feature of the process; it allows the layout of plastic regions, rigid regions and velocity discontinuities to be determined by inspection of the FELA mesh. The corresponding slip-line field can then be constructed numerically in the usual way. The examples presented are restricted to purely cohesive soil, but the same approach is applicable in principle to frictional or cohesive-frictional materials.
KEYWORDS: limit state design/analysis; numerical modelling; plasticity
ICE Publishing: all rights reserved

INTRODUCTION Many problems of limit analysis in plane strain can be solved very efficiently and potentially exactly by constructing the relevant slip-line field. As noted by Hill (1950):
The equations of plane strain are hyperbolic, and the characteristics are the slip-lines. This property compels us, if we wish to solve special problems, to treat the field of slip-lines as the fundamental unknown element to be determined.

The aim of this paper is to show that a range of previously challenging geotechnical stability problems can now be solved by using finite-element limit analysis (FELA) primarily as a tool for revealing the slip-line field, rather than as a means of obtaining direct lower and upper bounds on the exact collapse load (though the bounds from FELA still provide a valuable check on the correctness of the inferred slip-line solution). For a given problem, the approach followed here proceeds in two stages. The first involves FELA using the authors program OxLim, which combines the lower- and upper-bound methods of Makrodimopoulos & Martin (2006, 2007, 2008) with a simple strategy for adaptive mesh refinement. Essentially, this strategy attempts to equalise : the quantity cmax dA (integral of the maximum shear strain rate) over all elements of the mesh, subject to the proviso that no element be de-refined from its initial size. After several cycles of adaptive refinement, the concentration of ele: ments begins to reflect the intensity of cmax . For a material that is both homogeneous and purely cohesive, the plastic work rate is directly proportional to the maximum shear strain rate, so in this case the adaptivity strategy may eqManuscript received 1 March 2011; first decision 24 March 2011; accepted 5 April 2011. Published online at www.geotechniqueletters.com on 11 May 2011. * Department of Engineering Science, University of Oxford, Oxford, UK

ually well be viewed as one that seeks to equalise : each elements contribution to the total internal work rate W int (Martin, 2009). In a highly refined mesh, velocity discontinuities appear as dark bands with numerous tiny elements, whereas rigid blocks remain unrefined (except as required to maintain element quality in the vicinity of an adjacent plastic region or velocity discontinuity). Regions of diffuse plastic shearing undergo intermediate levels of refinement. The generation of unstructured triangular meshes in OxLim is performed using the program Triangle (Shewchuk, 2002). The second stage of the solution process involves construction and verification of the slip-line field. These calculations are performed numerically, using standard finite-difference procedures derived from the method of characteristics (e.g. Hill, 1950; Houlsby & Wroth, 1982; Sokolovskii, 1965). Within a plastic region, the calculated stress field automatically satisfies both equilibrium and the yield criterion. If any rigid blocks are present, however, additional constraints must be imposed to ensure that they too are in equilibrium. The calculated velocity field within a plastic region is automatically consistent with the stress field in the sense that the principal strain directions are certain to be aligned with the principal stress directions. It is possible, however, for the two sets of directions to become mismatched in some areas, such that the major principal stress is aligned with the minor principal strain and vice versa. For a purely cohesive soil (i.e. undrained analysis with c 5 su and Q 5 0) there is actually no need to check for this occurrence because the associated flow rule remains satisfied regardless; any inconsistency will manifest itself as a difference between the lower and upper bounds that are calculated from the slip-line stress and velocity fields, respectively. A slip-line field is generally set up as a function of various geometrical parameters (angles, lengths, radii, etc.) that need to be adjusted until all boundary conditions and rigid block equilibrium constraints, if any are satisfied for the problem at hand. For non-trivial problems this requires the numerical solution of n non-linear equations in n unknowns, as discussed by Martin (2005) in the context of 23

24

Martin Repeating the process for C/D 5 3 and cD/su 5 3, OxLim gives 26?158 # N # 26?079 and the mesh shown in Fig. 3(a). It is clear that the collapse mechanism is now more deep-seated, with the whole tunnel being surrounded by plastically deforming soil. Close inspection also reveals that two velocity discontinuities intersect near the tunnel invert. Nevertheless, it is fairly straightforward to construct the slip-line field, which is shown in Fig. 3(b) and gives N 5 26?085. This is consistent with the bounds from OxLim, and it is also within the range 26?49 # N # 25?63 obtained by Sloan & Assadi (1993). Figure 4 compares the FELA and slip-line collapse mechanisms; the agreement is again convincing. It is interesting to note that these slip-line fields have much in common with the solution recently identified for the trapdoor problem (Martin, 2009). STRIP FOOTING UNDER ECCENTRIC LOADING Various solutions for the undrained bearing capacity of a rigid strip footing under combined vertical, horizontal and moment (V, H, M) loading are available (see, for example, Salenc on & Pecker (1995), Bransby & Randolph (1998) and Ukritchon et al. (1998)). The (V, H) interaction can be determined analytically (Green, 1954), but much less is known about the (V, M) interaction and the associated slip-line fields. Only two levels of vertical load are considered here: V/V0 5 0?5 and 0?75, where (V0 5 (2 + p)Bsu and B is the width of the footing). For simplicity, it is assumed that the soilfooting interface can sustain unlimited tension. When V/V0 5 0?5, OxLim gives 0?6689 # M/B2su # 0?6815 and the adaptively refined mesh in Fig. 5(a) reveals an interesting variation on the standard scoop mechanism. The slip-line field, shown in Fig. 5(b), is easy to construct and gives M/B2su 5 0?6749. It has been verified that the soil at the left-hand edge of the footing is not overstressed with respect to the free surface, so there is no obvious barrier to extension of the stress field. Figure 6 confirms the expected agreement between the FELA and slip-line collapse mechanisms. When V/V0 5 0?75, the bounds from OxLim are 0?5131 # M/B2su # 0?5216, and the mesh in Fig. 7(a) shows that the collapse mechanism is surprisingly complicated. In particular, there are now two rigid blocks meeting at a hinge point; there is also a band of diffuse shear emanating from the fan region and propagating to the left, rather than a single slip-line (with velocity discontinuity) in the shape of a circular arc. The slip-line field, shown in Fig. 7(b), gives M/B2su 5 0?5187. The FELA and slip-line collapse mechanisms are compared in Fig. 8. In this instance the agreement is particularly compelling because of the complexity of the mechanism. VERTICAL CUT The final example is one of the fundamental problems of geotechnical stability analysis: the collapse of a vertical cut of height H in a homogeneous soil with unit weight c and undrained strength su. The exact value of the stability number N~ cH su (2)

bearing capacity. In this study, the equation system is solved using the hybrid algorithm of Powell (1970). It is worth pointing out, however, that several of the slip-line fields in the examples given are not statically determinate, meaning that both the stress field and (part of) the velocity field need to be constructed at each iteration of the solver. If the adjusted slip-line field: (a) gives a collapse load that falls between the FELA lower and upper bounds, and (b) gives a collapse mechanism that matches the one obtained from FELA there is a strong indication that the correct slip-line field has been identified. Of course, it cannot be stated with absolute certainty that the exact solution has been found until the slip-line stress field has been extended into all rigid regions in a manner that is both statically and plastically admissible. This is beyond the scope of the present paper, but it has been verified that the slip-line stress fields given later are all locally extensible; that is, there are no salient points where elements of rigid soil are overstressed with respect to adjacent slip-lines or boundaries (see Hill (1954) for further details of the checks involved).

SHALLOW CIRCULAR TUNNEL Plasticity solutions for the collapse of circular tunnels have been developed by several researchers, notably Davis et al. (1980) and Sloan & Assadi (1993). For homogeneous undrained soil, the problem is usually posed in terms of the parameters C/D (cover to diameter ratio) and cD/su (dimensionless self-weight to strength ratio), with the result expressed in terms of the stability number N~ ss {st su (1)

where ss is the surcharge pressure acting on the ground surface and st is the support pressure inside the tunnel. Two illustrative cases are considered here. When C/D 5 3 and cD/su 5 0 (i.e. the academic case of a weightless soil), OxLim gives 4?063 # N # 4?140 using its default bracketing target of 1%. Some further adaptive refinement (to reveal the arrangement of plastic regions, rigid blocks and velocity discontinuities with greater clarity) gives the mesh shown in Fig. 1(a). The slipline field can immediately be recognised and constructed as shown in Fig. 1(b). Note that in these and all subsequent diagrams, the coordinates are normalised by the relevant dimension of the problem (in this case the tunnel diameter), crosses are used to mark centres of rotation and heavy lines denote velocity discontinuities. As the subdivision counts in the slip-line analysis are systematically increased, the calculated stability number converges to N 5 4?132. As expected, this lies between the OxLim bounds, and it is also within the range 3?78 # N # 4?51 obtained by Sloan & Assadi (1993). The FELA and slip-line collapse mechanisms are compared in Fig. 2. The close agreement, although qualitative, gives additional confidence that the slip-line field has been identified correctly. To give some idea of the run times involved, on a 3 GHz computer it took OxLim a total of 10 s to bracket N to 1%, and 154 s to reach the level of mesh refinement shown in Fig. 1(a). The slip-line calculations, starting with Fig. 1(b) and using successive doublings of the subdivision counts, required less than 1 s on a 2?4 GHz computer to obtain N converged to four-digit precision.

has so far defied solution, but the latest bounds from largescale FELA (Kammoun et al., 2010; Pastor et al., 2009) are 3:77522 N 3:77756 (3)

The use of adaptive finite-element limit analysis to reveal slip-line fields

25

_1

_1

_2

_2

_3

_3

_4

_4

_5

1 (a)

_5

1 (b)

Fig. 1. Tunnel collapse with C/D 5 3 and cD/su 5 0: (a) adaptively refined mesh from FELA; (b) slip-line field

_1

_1

_2

_2

_3

_3

_4

_4

_5

1 (a)

_5

1 (b)

Fig. 2. Tunnel collapse with C/D 5 3 and cD/su 5 0: comparison of collapse mechanisms from (a) FELA and (b) slip-line solution

Figure 9(a) shows an adaptively refined mesh for this problem obtained using OxLim. The picture reveals a curious arrangement of plastic regions, along with two rigid blocks that meet at a hinge point. It is also apparent that the velocity discontinuity along the right-hand

perimeter of the mechanism is the only one occurring. The slip-line field is shown in Fig. 9(b), and there is a persuasive resemblance to Fig. 9(a). The FELA and slipline collapse mechanisms, shown in Fig. 10, also agree very well. Note the slight clockwise rotation of the large

26

Martin

_1

_1

_2

_2

_3

_3

_4

_4

_5

1 (a)

_5

1 (b)

Fig. 3. Tunnel collapse with C/D 5 3 and cD/su 5 3: (a) adaptively refined mesh from FELA; (b) slip-line field

_1

_1

_2

_2

_3

_3

_4

_4

_5

1 (a)

_5

1 (b)

Fig. 4. Tunnel collapse with C/D 5 3 and cD/su 5 3: comparison of collapse mechanisms from (a) FELA and (b) slip-line solution

rigid block that is visible in both Fig. 10(a) and Fig. 10(b). Table 1 shows that as the slip-line subdivision counts are successively doubled, the calculated stability number converges to

N ~3:77649 . . .

(4)

The final analysis took less than 2 min on a 2?4 GHz computer. As expected, the slip-line stability factor falls within the narrow range of equation (3). Although it seems

The use of adaptive finite-element limit analysis to reveal slip-line fields


0.5 0.5

27

0.0

0.0

_0.5

_0.5

_1.0 _1.5

_1.0

_0.5

0.0 (a)

0.5

1.0

1.5

_1.0 _1.5

_1.0

_0.5

0.0 (b)

0.5

1.0

1.5

Fig. 5. Strip footing moment capacity when V/V0 5 0?5: (a) adaptively refined mesh from FELA; (b) slip-line field
0.5 0.5

0.0

0.0

_0.5

_0.5

_1.0 _1.5

_1.0

_0.5

0.0 (a)

0.5

1.0

1.5

_1.0 _1.5

_1.0

_0.5

0.0 (b)

0.5

1.0

1.5

Fig. 6. Strip footing moment capacity when V/V0 5 0?5: comparison of collapse mechanisms from (a) FELA and (b) slip-line solution
0.5 0.5

0.0

0.0

_0.5

_0.5

_1.0 _1.5

_1.0

_0.5

0.0 (a)

0.5

1.0

1.5

_1.0 _1.5

_1.0

_0.5

0.0 (b)

0.5

1.0

1.5

Fig. 7. Strip footing moment capacity when V/V0 5 0?75: (a) adaptively refined mesh from FELA; (b) slip-line field

0.5

0.5

0.0

0.0

_0.5

_0.5

_1.0 _1.5

_1.0

_0.5

0.0 (a)

0.5

1.0

1.5

_1.0 _1.5

_1.0

_0.5

0.0 (b)

0.5

1.0

1.5

Fig. 8. Strip footing moment capacity when V/V0 5 0?75: comparison of collapse mechanisms from (a) FELA and (b) slip-line solution

28

Martin

1.0

1.0

0.5

0.5

0.0

0.0

Note: centres of rotation are at (1.42,0.52) and (_2.91,3.24), out of range of axes

0.0

0.5 (a)

1.0

0.0

0.5 (b)

1.0

Fig. 9. Collapse of vertical cut: (a) adaptively refined mesh from FELA; (b) slip-line field

1.0

1.0

0.5

0.5

0.0

0.0

Note: centres of rotation are at (1.42,0.52) and (_2.91,3.24), out of range of axes

0.0

0.5 (a)

1.0

0.0

0.5 (b)

1.0

Fig. 10. Collapse of vertical cut: comparison of collapse mechanisms from (a) FELA and (b) slip-line solution

highly likely that the exact solution has now been found, at least to six-digit precision, formal confirmation will require the stress field to be extended into the two rigid blocks within the mechanism, as well as the semi-infinite soil mass outside. This could well prove to be a difficult task, but preliminary checks have confirmed that there are no problems with extensibility at critical locations. In particular, the soil is not overstressed at the hinge point, nor is it overstressed at the toe of the cut. CONCLUSIONS This paper has shown how FELA, when combined with a simple strategy for adaptive mesh refinement, can be used

to reveal the arrangement of plastic regions, rigid blocks and velocity discontinuities in a range of geotechnical stability problems in plane strain. Using the FELA program OxLim, the resolution that can be achieved is so sharp that the relevant slip-line field can readily be deduced from inspection, and then constructed in a separate analysis using the method of characteristics. The procedure has been demonstrated for a few selected problems of undrained stability, but the same strategy should also be applicable to problems involving frictional or cohesivefrictional materials. The new approach shows great promise for obtaining exact solutions to all manner of plane strain plasticity problems where it has, to date, proved impossible to identify the correct slip-line field by trial and error.

The use of adaptive finite-element limit analysis to reveal slip-line fields


Table 1. Convergence of lower- and upper-bound stability factors calculated from slip-line field for vertical cut Slip-line subdivision counts with reference to Fig. 9(b) 61 62 64 68 616 632 Lower bound (from stress field) 3?7759789 3?7763630 3?7764590 3?7764829 3?7764889 3?7764904 Upper bound (from velocity field) 3?7766116 3?7765205 3?7764983 3?7764928 3?7764914 3?7764911

29

REFERENCES
Bransby, M. F. & Randolph, M. F. (1998). Combined loading of skirted foundations. Ge otechnique 48, No. 5, 637655. Davis, E. H., Gunn, M. J., Mair, R. J. & Seneviratne, H. N. (1980). The stability of shallow tunnels and underground openings in cohesive material. Ge otechnique 30, No. 4, 397416. Green, A. P. (1954). The plastic yielding of metal junctions due to combined shear and pressure. J. Mech. Phys. Solids 2, No. 3, 197211. Hill, R. (1950). The mathematical theory of plasticity. Oxford: Clarendon. Hill, R. (1954). On the limits set by plastic yielding to the intensity of singularities of stress. J. Mech. Phys. Solids 2, No. 4, 278 285. Houlsby, G. T. & Wroth, C. P. (1982). Direct solution of plasticity problems in soils by the method of characteristics. Proc. 4th Int. Conf. on Numerical Methods in Geomechanics, Edmonton 3, 10591071. Kammoun, Z., Pastor, F., Smaoui, H. & Pastor, J. (2010). Large static problem in numerical limit analysis: a decomposition approach. Int. J. Numer. Anal. Methods Geomech. 34, No. 18, 19601980. Makrodimopoulos, A. & Martin, C. M. (2006). Lower bound

limit analysis of cohesive-frictional materials using secondorder cone programming. Int. J. Numer. Methods Engng 66, No. 4, 604634. Makrodimopoulos, A. & Martin, C. M. (2007). Upper bound limit analysis using simplex strain elements and second-order cone programming. Int. J. Numer. Anal. Methods Geomech. 31, No. 6, 835865. Makrodimopoulos, A. & Martin, C. M. (2008). Upper bound limit analysis using discontinuous quadratic displacement fields. Commun. Numer. Methods Engng 24, No. 11, 911927. Martin, C. M. (2005). Exact bearing capacity calculations using the method of characteristics. Proc. 11th Int. Conf. on Computer Methods and Advances in Geomechanics, Turin 4, 441450. Martin, C. M. (2009). Undrained collapse of a shallow planestrain trapdoor. Ge otechnique 59, No. 10, 855863. Pastor, F., Loute, E. & Pastor, J. (2009). Limit analysis and convex programming: a decomposition approach of the kinematic mixed method. Int. J. Numer. Methods Engng 78, No. 3, 254274. Powell, M. J. D. (1970). A hybrid method for nonlinear algebraic equations. In Numerical methods for nonlinear algebraic equations (Rabinowitz P. (ed.)), 87114. London: Gordon and Breach. Salenc on, J. & Pecker, A. (1995). Ultimate bearing capacity of shallow foundations under inclined and eccentric loads. Part I: purely cohesive soil. Eur. J. Mech. A Solids 14, No. 3, 349375. Shewchuk, J. R. (2002). Delaunay refinement algorithms for triangular mesh generation. Comput. Geom. 22, No. 13, 21 74. Sloan, S. W. & Assadi, A. (1993). Stability of shallow tunnels in soft ground. In Predictive soil mechanics (Houlsby G. T. & Schofield A. N. (eds)), 644663. London: Thomas Telford. Sokolovskii, V. V. (1965). Statics of granular media. Oxford: Pergamon. Ukritchon, B., Whittle, A. J. & Sloan, S. W. (1998). Undrained limit analyses for combined loading of strip footings on clay. J. Geotech. Geoenv. Engng 124, No. 3, 265276.

WHAT DO YOU THINK?

To discuss this paper, please email up to 500 words to the editor at journals@ice.org.uk. Your contribution will be forwarded to the author(s) for a reply and, if considered appropriate by the editorial panel, will be published as a discussion.

Você também pode gostar