PROCEEDINGS OF THE SEVENTH INTERNATIONAL CONFERENCE ON COMPUTER, METHODS AND ADVANCES IN GEOMECHANICS / CAIRNS /6- 10 MAY 1991 Computer Methods and Advances in Geomechanics Edited by G.BEER CSIRO, Division of Geomechanics J.R.BOOKER & J.RCARTER University of Sydney OFFPRINT Ns Published on behalf of the International Association for Computer Methods and Advances in Geomechanics by A.A.BALKEMA / ROTTERDAM / BROOKFIELD / 1991Computer Methods and Advances in Geomechanics, Beer, Booker & Carter (eds) ‘© 1991 Balkema, Fotterdam. ISBN 906191 1893, Non-linear finite element analysis of safety factors R.B.J.Brinkgreve Delft University of Technology, Netherlands HLL. Bakker Public Works Department, Netherlands ABSTRACT: In this paper a robust method is proposed to determine the safety factor of geotechnical constructions in finite element computations. The method 1s based on the SeSuction of the strength parameters of soil, the friction angle # and the cohesion c. Three examples show the practical application of the method. 1 INTRODUCTION In structural engineering a factor of safety 1s always defined as the ratio of the collapse load over the the working load. The same definition is adopted in foundation engineering, at least for footings and piles. For soil bodies such as road/river enbanknents and earthen dams the situation is different. Here the dominating load is not a direct external force, but most of the load cones from soil weight Very cohesive soil bodies can be loaded to collapse by increasing gravity, either numerically or in centrifuge tests, but not when the strength 1s dominated by friction. Therefore other definitions of safety are connon in soll mechanics. The usual soil mechanics definition of safety is: S _ eto'tan safety factor = -§-* “ES er tane, where S is the shear strength, further defined by the well known Mohr-Coulomb criterion, Note that c is the cohesion, # is the friction angle and o’ is the normal pressure at the plane considered. Both c and ¢ are effective strength parameters and o” is an effective stress. We use the subscript ¢ to indicate critical strength paraneters; Just high enough to ensure equilibrium. Indeed, for c=c, and g=p_ the safety factor becomes equal to unity. The above definition of safety coincides with the classical definition as used in slip circle analysis on the condition that we define: 1417, ef tang, = ae / tang with ast In this study the proportionality with will be retained in order to remain compatible to traditional slip circle analysis, but the method allows for other ratios of c. and tané,. ‘This study concentrates on the computation of the above safety factor by use of an elastic-plastic finite element method. In ‘stead of the usual incrementation of loads, strength parameters will be decremented. This technique was first proposed by Zienkiewicz at al. (1975), but we make the Procedure robust by adding an arc-length . technique. A robust procedure was needed for implementing into the PLAXIS finite element package. Finally the potential of ‘the method is demonstrated by considering a number of applications. 2 MOHR-COULOMB MODEL, Being interested in collapse loads rather than precise deformations, the elastic-plastic Mohr-Coulomb model 1s adopted. Hence x 1 oy, B adgeep eeeThe symbols f and g are used to denote the yleld function and the plastic potential function respectively. Considering planar deformations only, with the x-y-coordinates in the plane and the 2-axis normal to the plane, we define fer-(ccotp+e) sing get-o sing where Ce (oto) we Sy’ 2a? yy! For the sake of convenience, it is assumed that the out-of-plane stress ois in between the principal stresses o-t and c+. 3 IMPLICIT INTEGRATION Finite element analyses involve finite increments of stress and strain rather than rates as considered above. Several integration rules can be applied to pass from rates to finite increments. One of the most popular ones, as first proposed by Vermeer (1979) and most recently by BorJa & Lee (1990), 1s the implicit integration schene. Here the direction of flow 1s considered to be fully determined by the state of stress at the end of a load Ancrenent. Denoting finite increnents by the symbol A and t for time, we get . 28 ay = 28 a8 as sa Mare Bsra- fo where f° = flo + D be) where @'ls the stress at the beginning of the load increment. Details are given by Vermeer & Van Langen (1989). Having defined the increment of plastic strain as a function of the total strain Increment, is easy to compute the stress increment from: at an D (ae ~Be?) 4 FINITE ELEMENT FORMULATION We consider a soil body subJect to constant gravity and constant external loads. These are equilibrated by a known stress field @°. Hence, the finite element equilibrium Conditions give: 1118 ¢ Fig. 1. Increnents for strength-reduction SBer a where both body forces and external loads are assembled in the vector Q for nodal forces. We now apply a decrement of the strength parameters, so that the yield locus rotates as indicated in Fig.1. Obviously, the new stresses ¢ (without superscript) will be at the new yield locus, at least for material points with stresses ¢° above this line. Let us now define = 9° bot bo® 9 4g7= D (Ae ~he") where 9°+ 4g" is the closest point projection of @° on the new yield surface. Substitution in the equillbrium equation, and using ¢= Bu, gives: K ou = RY AR + P where K is the elastic stiffness matrix and 1, aR =~ tg" Jy BD Me? av In practice R° will not entirely vanish because of numerical procedures; separation fron AR 1s important for later formulation of an arc-length control procedure. P is a pseudo load vector, ae used within an initial-stress iteration-procedure of the We adopt an initial stress procedure since a tangent stiffness procedures tend to break down in the fully plastic range due to 111-conditioning.en 5 INDIRECT DISPLACEMENT CONTROL, The above strength reduction procedure is is far from being robust. Indeed, by a step wise decrease of strength it 1s possible to approach collapse, but for some step the strength will be reduced too much, i.e. beyond critical, so that the iterative procedure will not converge to an equilibrium state of stress. In this manner the precise critical values of tang and c are never obtained. Therefore indirect displacement control is needed, or in other words arc-length control (Rheinboldt & Riks, 1986). Within this approach we add a load multiplier 8, and obviously an extra scalar equation for solving B. The system of iteration equations is K out = Res ata + PR aut ast = au? au, The subscript 1s used to denote results fron the previous load step. Instead of the node displacement vector of the previous step, other vectors can be used to obtain a Linear scalar equation for 8. Similarly non-linear equations are feasible. 6 STABILITY OF RIVER EMBANKMENT ‘The first practical application involves a river embanknent in the tidal zone. The slope has a horizontal length of 10 m and a height of Sm, so the ratio is 1:2. The most dangerous situation for a embanknent 4s at the onset of low tide. The waterlevel outside the embankment 1s already low, but the phreatic level in the embankment is still high. The situation, which is modelled 1s shown in Fig. 2. We consider a homogeneous soll with material properties as Indicated in Table 1. Fig. 2. River embankment at low tide 1119 Table 1. Soll properties for embankment Parameter ‘Symbol | Value Dry weight % 16.0 KN/n? Wet weight 1 20.0 kN/m? Friction angle ’ 30.0 ° Cohesion e 5.0 kPa Shear modulus G 1000 KPa Poisson’s ratio | v 0.3 In the first part of the analysis we apply the soil weight to introduce the stresses for the situation of Fig. 2. Some of the Gaussian integration points appear to be in plastic state already, but the embankment has not collapsed at all In the second part of the calculation we want to analyze the safety of the construction. Therefore we apply ten steps in which the strength parameters of the soil are stepwise reduced, according to the theory as described before. When further reduction of the strength parameters in not possible anymore, the construction has collapsed and the safety factor is obtained. Fig. 3 shows the computed strength-displacement curve for the crest point of the embankment The value of the safety factor for this particular problem appears to be 1.20, strength reduction Diaplacenents crest point * 0.1 Fig. 3. Strength-displacement curve Fig. 4 shows a plot of the incremental ‘displacenent contours at failure. In this plot one can recognize a nice slip circle. In the same plot the critical slip circle according to Eishop’s method is plotted, Note that we performed a drained analysis both for the finite element approach and the Bishop approach.Fig. 4. Incremental displacement contours Whilst the positions of the slip circles are sonevhat different, the safety factor in both Bishop's method and the finite elenent calculation 1s exactly the sane. This gives sone validation of the proposed method. It might now be concluded that the calculation of the safety factor in finite element codes has no advantage in comparison with conventional methods. In this simple case this happens to be true, but the finite element method will also detect non-circular slip surfaces and associated factors of safety. The next examples will show that in general the finite elenent calculation of the safety factor gives information that cannot be obtained with conventional methods. 7 STABILITY OF A SHEET-PILE WALL For the second application we consider the safety of an excavated building trench. The excavation depth is 21 m. The soll is supported by 32 m long sheet-pile walls, which are strutted at the top. During the excavation of the soil the water table remains at a constant level, both inside and outside the trench. Fig. 5 shows the finite element mesh of the trench. Special Line elements, as available in the PLAXIS Fig. 5. Mesh for building trench sheet-pile wall (Bakker & Brinkgreve, 1990). The dashed lines in Fig. 5 Andicate Joint elements to model reduced friction at the soil-structure interface (Van Langen & Vermeer, 1990). The soil is mostly fine silty sand with sone clay layers. In fact it consists of 7 layers, which correspond to the element layers in the mesh of Fig 5. The material properties for each layer are given in Table 2. Layer No. 1 is the top layer and layer No. 7 is the lowest layer. For the sheet-pile wall and the strut ve used: Tynest = 1:65 x 10° Nn? A F lanes, 7 8:04 x 10° kN/m erat’ b= 1:28 x 10% kN/a? : Table 2. Properties for different layers %|% | @/[e] ¢ Paver) ml ed] wave) Na wpa | KPa 1 | 18.0 | 20.0 | 30.0 [1.0 | 6000 2 | 16.0 | 16.0 | 25.0 | 8.0] 3000 3 20.0 | 32.5 | 0.0 | 6000 4 20,0 | 32.5 | 0.0 | sooo | | 5 20.0 | 32.5 | 0.0 | 25000 | } 6 18.0 | 27.5 | 4.0 | 9500 7 20.0 | 32.5 | 0.0 | 15000 The soil which is to be excavated, is modelled by external loads: horizontal tractions simulating the horizontal stresses at the sheet-pile wall and vertical tractions simulating the vertical stresses at the bottom of the trench. The calculation consists of 3 parts. The first part 1s the introduction of the initial stresses including the tractions on the sheet-pile wall and the trench bottom. ‘The second part is the excavation of the trench by stepwise removing the external loads. Bending moments and strut forces occur in this part of the calculation. There is a stress reduction directly behind * the sheet-pile wall mainly due to arching (Bakker & Brinkgreve, 1990). The extrene horizontal displacement of the sheet-pile wall 1s about 100 mn. The extreme bending moment amounts 2100 kNa/m and the strut force is 500 kN/m. Fig. 6 shows a plot of the incremental displacement contours at service state. A concentration of contour Lines can be regarded as the initialisation of a shear band, 1120Fig. 6. Velocity contours at service state The third part of the calculation is the determination of the safety factor by means of strength reduction. According to this calculation, the safety factor appears to be 1.65. The development of the bending moments during this calculation part is remarkable, as the clasping moment disappears and the sheet-pile wall tends t behave like a beam on two supports (Bakker & Brinkgreve, 1991). Fig. 7 shows a plot of the incremental displacenent contours at collapse. This plot visualizes a slip line going from the right-hand top, underneath the sheet-pile wall to the middle of the trench bottom. I comparison with the service state the slip Line has moved drastically downwards. Fig. 7. Velocity contours at failure ‘8 EMBANKMENT STABILITY WITH GEOTEXTILE The third application involves the safety of a road embankment. The embankment has a height of 6 m. In this example two situations are modelled: Firstly an Fig. 8. Geometry of road embanknent ordinary case where the embankment is build Just on top of the underground and secondly With the addition of a geotextile underneath the embankment. The situation is shown in Fig. 8, The so11 consists of four layers. The top layer is the embankment fill material. The underground consists of two clay layers with a layer of soft peat in between. The Soil properties are given in Table 3. Table 3. Properties for different layers ealccaleon cae al ae oe % pavn?hrave?] + | wre | ue 1 |rs.0 [isc fas.o | 2.0 | soo a | 2 liste fisce fac | £0 | 500 Jo.co 3 fitce lire [isco | 2:0 | aso fo-rs 4 fisce [tae [2ao | 5:0 | 800 foceo Again the analysis consists of 3 parts. The first part 1s the introduction of the initial stresses in the underground. As the underground is horizontal the initial stresses can directly be derived from the soil weight and the K,-values given in Table 3. The second part of the calculation is the building of the embankment. At the end of this part the top of the embankment shows a settlement of 68 cm (without geotextile) and 66 cm (with geotextile). The geotextile seens to have Little influence so far. ‘The third part is the calculation of the safety factor. Now the differences between both situations are much more apparent. In the case without geotextile the safety factor 1s less than 1.20 while in the case with geotextile the safety factor is over 1.30. In practice, only the latter situation 1s acceptable. 1124The difference in safety factor between the two situations is due to the fact that the failure mechanisms are completely different. Fig. 9 and 10 show plots of the incremental displacement contours of both situations. In the first case the failure nechanism is a slip circle going through the soft peat layer. In the second case the same mechanism is initially developed, but this is resisted by the geotextile. Collapse finally occurs in the embanknent itself. Fig.9. Velocity contours without geotextile Fig. 10.Velocity contours with geotextile Considering the results it would seem that geotextiles have an effect on stability rather than on deformation, but this will depend on the relative stiffness of the geotextile. For the present application we used EA = 2500 kN/n. ‘9 CONCLUDING REMARKS: The proposed finite element method for the determination of the safety factor by means of strength reduction gives results which agree well with those from Bishop’ s slip-circle method, at least for simple situations when a circular surface occurs. The advantage of the method in comparison with conventional methods is that it even deals with the most complex kind of geotechnical constructions. Besides one does not have to predefine the failure mechanism. Arc-length control makes the procedure robust since failure need not be associat with a non-converging iterative procedure. Moreover the arc-length procedure gives th full strength-displacenent curves beyond possible peaks. REFERENCES Bakker, K.J. & Brinkgreve, R.B.J. 1990. The, use of hybrid beam elements to model sheet-pile behaviour in two dimensional deformation analysis. In Proc. II Eur. ‘Spec. Conf. on Num. Meth. in Geotech. Eng., p-559-571. Santander, CEDEX Bakker, K.J. & Brinkgreve, R.B.J. 1991. Deformation analysis of a sheet-pile wall, using a two dimensional model. In Proc. X Eur. Conf. on Soil Mech. and Foundation Eng., Florence Bor Ja, R.I. & Lee, S.R. 1990. Cam-Clay plasticity, Part I: Implicit integration of elasto-plastic constitutive relations. In Comp. Meth. in Appl. Mech. and Eng, Vol 78, p.49-72 Rheinboldt, W.C. & Riks, E, 1986. Solution techniques for non-linear finite elenents equations. In State-of-the-art surveys on finite element techniques, p. 183-223. New York, Appl. Mech. Div. of ASME Van Langen, H. & Vermeer, P.A. 1990. Finite element analysis of a pile penetration problem in clay. In Proc. IT Eur. Spec. Conf. on Num. Meth. in Geotech. Eng., p.519-527., Santander, CEDEX Vermeer, P.A. 1979. A modified initial strain method for plasticity problems. In Proc. III Int. Conf. on Num. Meth. in Geonech., p.377-387. Rotterdam, Balkema Vermeer, P.A. & Van Langen, H. 1989. Soil collapse computations with finite elements. In Ingenieur-Archiv 59, p. 221-236 Zienkiewicz, 0.C., Humpheson, C. & Lewis, R.W. 1975. Associated and non-associated visco-plasticity and plasticity in soll mechanics. Géotechnique 25, No. 4, p.671-689)Beer, G.,J.R.Booker & J.PCarter(eds.) 906191 1893, ‘Computer methods and advanees in geomechanics ~ Proceed- ings of the seventh international conference, Cairns, 6— 10 May 1991 1991, 25 m,¢.2000 pp., 3 vols., H.385 /$210.00/£122 ‘Computer methods have become a powerful tool for solving prob- lems in geomechanics & geotechnology. For example, in mining ‘civil engineering these methods have lead to improvements in design and are now applied to such diverse problems as rock ‘mechanics, geophysics, geological engineering, ice mechanics, blasting & environmental geotechnology. The proceedings present the latest intemational research onthe application of numerical 1 thods to problems in geomechanics & on advances in constitu- tive modelling of geomaterials. In addition, new measurement & monitoring techniques & analytical methods ae presented. Topics range fro. environmental to resource geotechnology including rock mechanics, geophysics, geological engineering, blasting, flow problems & more. Significant contributions are included on ‘CADiexper systems, groundwater geomechanics, back analysis experimental studies. FROM THE SAME PUBLISHER: 906191 809 X in geomechanics: Innsbruck 1988 Proceedings ofthe sith international conference, Innsbruck, 1! = 1S April 1988 1988-89, 25 cm, 2378 pp.,4 vols, H.450/$250.00 /£142 Main lectures; Numerical techniques & programming tive laws of geotechnical materials; Flow & consolidation; Ice ‘mechanics; Rock hydraulics; Modeling of joints, interfaces & discontinuumm; Modeling of infinite domains; Soil-structureinter- action ples: Earth structures, slopes, dams, embankments; Tun nels & underground openings; Dynamic & earthquake engineer- ing problems, blasting: Mining applications; Interpretation of field icrocomputers; Cad, mesh gene- Dungar,R. & J.Studer (eds.) 906191518X ‘Geomechanical modelling in engineering practice 1986, 25 cm, 409 pp, HA.185 /$100.00 /£58 ‘The Key to successful solution of problems by the finite element ‘method lies in the choice of appropriate numerical mod- ‘ls & their associated parameters for geological media. 16 invited contributions from wellknown authorities from USA, UK, Swit zerland, Japan && Canada on: Basic concepts; Nu ‘merical modelling of selected engineering problems; Specific mu- ‘merical models & parameters evaluation. Kotkman, PA.,J.Lindenberg & K.W.Pilarczyk (eds.) 906191 8154 Modelling soil-water.structure interactions - SOWAS 88— Proceedings ofthe international symposium. Delt 29.08-02.09.1988 1988, 25 cm, 514 pp., HA.150/$85,00/£47 Soil-water-structure interactions; Wave and current induced ‘behaviour ofthe seabottom: Local scour; Behaviour and stability ‘of block revetments and filter layers; Wave impact loads and ‘behaviour of asphalt revetments; Pils, platforms, piers and grav- ity structures; Sand suppletion processes and flow slides; Break- waters, dams and walls. Miscellaneous. Balasubramaniam, A.S..etal (eds) 906191 8642 Computer and physical modelling in geotechnical engineering Proceedings ofthe international symposium, Bangkok, 3-6 De- cemiber 1986 1989, 5 em, 550 pp. HA.190/$105.00/£60 Stability of natural & man made slopes: Design & design of foundations; Underground openings & excavations: Computer controled testing & investigation of soils; Data acquisition & ‘management in geotechnical engincering; Computer aided solu- tions for some special problems in engineering. 47 papers. Ervin, MC. (ed) 906191 5066 In-situ testing for geotechnical investigations - Extension course, Sydney May-June 1983 1983, 25 cm, 140 pp., HA.100/$55,00/£32 ‘An introduction to new developments, methods & applications of in-situ testing. 9 papers by Australian practising engineers & uni- versity lecturers. Editor: Coffey & Parmers, Melbourne. ‘Alemayehu Teferra & Edgar Schultze 906191 8049 Formulae, charts and tables in the areas of soil mechanics and foundation engineering- Stresses in soils Formeln, Tafeln und Tabellen aus dem Gebiet Grundbau und Bodenmechanik — Bodenspannungen 1988, 25 cm, 300 pp.,HA.95 /$50.00/£30 ‘This handbook is the result of many years of arduous work of | systematically collecting the scattered scientific literature that is highly valuable to students and practicing engineer in the fields of soil mechanics and foundation engineering. The merit ofthe hand- ‘book lies not oly inthe compilation of important information that isnormally not easily accessible to students and practicing en- gineers, but also in the provision of adequate illustrative examples ‘that will facilitate the application of te formulae, charts and tables. n onder to give the user the additional possibilty torefer to the original works, references are given atthe end ofeach topic. Elastic-sotropic half-space (Surface; Loads in half-space); Two layers Vertical surface loading (Point loads; Line loads; Strip loading; Uniformly distributed rectangular loads; Axial symme- tric loads; Loading with any shape; Stresses at interface), Multiple layers (Stresses at interfaces; Circular uniform loading). Authors: ‘Addis Ababa University & German Academic Exchange, Joshi, R.C. & F.Griffiths (eds.) 9061917077 Prediction and performance in geotechnical engineering — Proceedings ofan international symposium on prediction and, performance in geotechnical engineering, Calgary, 17-19 June 1987 1987, 25 cm, 464 pp.,H.150/ $85.00 /£47 Invited papers; Prediction & performance of pile foundations; Soil improvement; General prediction & performance of soils, unique soils, retaining structure behaviour & tunnels; Environmental geo- technology: Triaxial testing of soils; Centrifuge model testing; Risk analysis. 51 papers Spencer, A.J.M. (ed) 906191 6828 ‘Continuum models of discrete systems ~ Proceedings ofthe fifth international symposium, Nottingham, 14=20 July 1985 1987, 25 cm, 248 pp., Hf, 145 /$80.00 /£46 Recent developments in the study of material behavior, with part- cular emphasis on the dificult but important problems of relating ‘behavior on the microscopic scale to that on the macroscopic scale. Authors were specially invited. 29 papers, All books available from your bookseller or directly from the publisher: AA Balkema Publishers, P.O.Box 1675, Rotterdam, Netherlands For USA & Canada: A.A, Balkema Publishers, Old Post Rd, Brookfield, VI, USA