Plaxis finite element code for soil and rock analyses

Plaxis Bulletin
issue 20 / October 2006

Modeling passive earth pressures on bridge abutments for nonlinear Seismic Soil Structure interaction using Plaxis
Prediction of soil deformations during excavation works for the renovation of Het Nieuwe Rijksmuseum in Amsterdam, The Netherlands Comparison of the effectiveness of Deep Soil Mix columns using 2-D and 3-D Plaxis

Colophon

Editorial New Developments

3 4 5

The Plaxis Bulletin is the combined magazine of Plaxis B.V. and the Plaxis Users Association (NL). The Bulletin focuses on the use of the finite element method in geotechnical engineering practise and includes articles on the practical application of the Plaxis programs, case studies and backgrounds on the models implemented in Plaxis. The Bulletin offers a platform where users of Plaxis can share ideas and experiences with each other. The editors welcome submission of papers for the Plaxis Bulletin that fall in any of these categories. The manuscript should preferably be submitted in an electronic format, formatted as plain text without formatting. It should include the title of the paper, the name(s) of the authors and contact information (preferably email) for the corresponding author(s). The main body of the article should be divided into appropriate sections and, if necessary, subsections. If any references are used, they should be listed at the end of the article. The author should ensure that the article is written clearly for ease of reading. In case figures are used in the text, it should be indicated where they should be placed approximately in the text. The figures themselves have to be supplied separately from the text in a common graphics format (e.g. tif, gif, png, jpg, wmf, cdr or eps formats are all acceptable). If bitmaps or scanned figures are used the author should ensure that they have a resolution of at least 300 dpi at the size they will be printed. The use of colour in figures is encouraged, as the Plaxis Bulletin is printed in full-colour. Any correspondence regarding the Plaxis Bulletin can be sent by email to bulletin@plaxis.nl or by regular mail to:

Results from the ERTC7 benchmark exercise

Plaxis Benchmark

Modeling passive earth pressures on bridge abutments for nonlinear Seismic Soil - Structure interaction using Plaxis Prediction of soil deformations during excavation works for the renovation of Het Nieuwe Rijksmuseum in Amsterdam, The Netherlands Comparison of the effectiveness of Deep Soil Mix colomns using 2-D and 3-D Plaxis

Plaxis Practice

Plaxis Practice 16

Plaxis Bulletin
c/o Erwin Beernink PO Box 572 2600 AN Delft The Netherlands The Plaxis Bulletin has a total circulation of 11.000 copies and is distributed worldwide. Editorial Board: Wout Broere Ronald Brinkgreve Erwin Beernink Franois Mathijssen

Plaxis Practice 20

Recent activities 23 Activities 2006-2007 24



Editorial

After some challenging activities during the past summer period including a number launches and user meetings we can proudly conclude that Plaxis Practice is still increasing world-wide. A spin-off of some of these activities can be found in this 20th issue of the Plaxis bulletin. This comprehensive issue contains very interesting papers by users about the contribution of Plaxis programs to everyday soil and rock anaylses. In the first paper Anoosh Shamsabadi and Steinar Nordal focus on the development of stiffness degradation due to earthquakes for the earth pressure behind bridge abutments. This problem has recently been studied at several universities and research institutes in the USA. Back calculations of the test results using Plaxis 3-D Tunnel and 3-D Foundations show that Plaxis is an extremely valuable tool for the practicing bridge engineer for capturing realistic nonlinear soil-abutment-structure interaction curves. The second contribution by H.D. Netzel and D. Vink of Crux Engineering discusses the geotechnical design calculations used in the recent renovation of the world famous Rijksmuseum in Amsterdam. Plaxis was used on this project. The calculations are part of the risk assessment strategy for predicting the influence of ground deformation due to excavation close to buildings. In the third paper Hester Leung, Cristien Gani, Wataru Okada and Sergei Terzaghi of SKM Auckland and SKM Sydney compare the behaviour of deep soil mixing columns predicted by Plaxis 2-D and 3-D programs. These columns are one of the remedial solutions for slip repair works on state highways in the Northland Region of New Zealand. Slope failure is quite common in this region due to heavy rainfall, particularly in winter months. 2-D and 3-D Plaxis were used as concerns were raised about the suitability of only two rows of columns in the design. This last paper shows the growing importance of 3D modelling. In recent years we have done a great deal of development work in this area. We are currently working on a general 3D finite element program in which different types of geotechnical applications with arbitrary geometries will eventually be combined. We hope that you will find this 20th edition instructive and enjoyable.

New Developments

New Developments
Ronald Brinkgreve, Plaxis BV

In recent years much development work has been spend on 3D modelling. In addition to 3D Tunnel, which was the first 3D program developed by Plaxis, most 3D developments by now have been performed in the Foundation program. And there is more to come. Plaxis is working on a general 3D finite element program in which, eventually, different types of geotechnical applications with arbitrary geometries can be combined, while maintaining the familiar easy-to-use modelling concepts. However, it will take some time before a first release of this general 3D program becomes available. Meanwhile, developments in the 3D Foundation program also continued and will make the program more versatile. The upcoming Version 2 will not only aim at the modelling of foundations (including multiple piles), but also more at excavations and slopes, although the name 3D Foundation is retained. In this contribution I will reveal some of its new features. In Bulletin 17 I already mentioned the embedded pile option based on the embedded beam concept [Ata.1998]. Although originally planned for last years release of 3D Foundation, we decided that this new technology would require more testing and validation, and therefore intermediate versions 1.5 and 1.6 were released without embedded piles. Nevertheless, a first implementation was ready then for research and testing purposes. Meanwhile, an extensive testing and validation programme is being carried out in cooperation with our network of contacts at universities and research institutes. Main focus of the embedded pile option is on axially loaded bored piles in compression and piles in extension. It is particularly useful for the optimisation of piled-raft foundations where the interplay between the raft, the piles and the soil is essential to make an economic design. In addition, we will also evaluate the possibilities for laterally loaded piles, although the embedded beam concept is not primarily meant for this type of applications. It is also not very suitable for displacement piles or driven piles, since it does not take into account the increase in lateral stress and the effects of densification in the soil around the pile. Nevertheless, the embedded pile option provides a valuable extension to the Foundation program. The embedded beam concept has also been combined with node-to-node anchors to model ground anchors in which the embedded beam acts as the grout body whereas the node-to-node anchor acts as the anchor rod. Ground anchors can be placed on work planes (at walls, beams or other structures) and can go arbitrarily into the ground. The anchor bearing capacity must be specified by the user in terms of a maximum friction force along the grout body. This feature is also being tested at the moment. Since Foundation version 1.5, the modelling of slopes and embankments has improved by means of the triangulate meshing option (to create straight slopes and sharp edges). Now, with version 2, the well-known phi-c reduction procedure for the calculation of global safety factors will also be available as a standard calculation option. Moreover, the new embedded pile and ground anchor can also be used to model reinforced slopes and embankments. Hence, version 2 further facilitates the modelling and analysis of these types of applications. For users of large 3D models that have been somewhat disappointed with the slow performance of the existing Output program, there is good news. The Output program has

been completely refactored and supports OpenGL (hardware accelerated graphics). This means that all graphic operations in the Output program are dramatically faster, provided that the computers graphic card has sufficient memory (>64 MB). Rotation and zooming into the model can be done using the mouse and scroll wheel, just like in popular CAD programs. Also the speed of output tables has been increased significantly, and several features have become more user-friendly. A particular feature to mention is the output of volume piles, which will be automatically provided in terms of structural forces, just as for the embedded piles, rather than as stress plots only. At the moment a full-function beta version of the new 3D Foundation V2 is being tested by a selected group of users. In the coming two months we will complete the testing programme and apply our Quality Assurance procedures for the final release of this new product, which is planned for January 2007.

References:
Ata N.1998. Etude du comportement des micropieux sous chargement lateral: Construction numerique des courbes (p-y) et coumplage fluide-squelette. PLD thesis, Lille University of Technology.

Plaxis Benchmark

Results from the ERTC7 benchmark exercise

H.F. Schweiger, Computational Geotechnics Group, Institute for Soil Mechanics and Foundation Engineering, Graz University of Technology, Graz, Austria

Abstract
A benchmark example addressing ultimate limit state (ULS) design of an embedded retaining wall has been specified by the ERTC7 on the occasion of the 6th European Conference on Numerical Methods in Geotechnical Engineering. The main goal of the exercise was to highlight possibilities and limitations of numerical methods for ULS design with particular reference to Eurocode7 (EC7), where three design approaches (DA1, DA2 and DA3) have been specified. These design approaches differ in the way partial factors of safety are introduced in the analysis. However, the exercise was not restricted to the use of EC7 and thus a wide spectrum of results, which are summarized in this contribution, could be expected. Some of the 13 submissions did not actually present design values as requested but provided parametric studies showing the influence of various design assumptions or did an analysis using characteristic values, some introduced factors without given explicit reference to a particular code or standard. rarely given in a conventional geotechnical report, is also left to the user in most cases and was therefore not given.

Specification of ERTC7 benchmark

The benchmark is a deep excavation problem supported by a single strut. The geometry is depicted in Figure 1. The significant difference to examples previously examined by various working groups around Europe (Orr, 2005) is that it was the intention here to solve this problem, including determination of the required embedment depth (!), by means of numerical methods, although a check by simple limit equilibrium calculations was certainly recommended. As the emphasis is on the ULS design and not on the serviceability limit state (SLS) only parameters required for simple elastic-perfectly plastic analysis have been provided (Table 1). These parameters have to be considered as characteristic values and not design values. The following construction steps should be modelled in the numerical analysis: - initial phase (K0 = 0.5) - activation of diaphragm wall (wished-in-place) - activation of surcharge loads - excavation step 1 to level -2.0 m - activation of strut at level -1.50 m, excavation step 2 to level -4.0 m, - groundwater lowering inside excavation to level -6.0 m - excavation step 3 to level -6.0 m The surcharge of 10 kPa is a permanent load, the surcharge of 50 kPa is a variable load. Bedrock was assumed at a depth of 20 m below ground surface. The axial stiffness of the strut was set to EA = 1.5E6 kN/m. Results to be provided: - Embedment depth of wall - Design bending moment for the wall - Design strut force On purpose it was not specified how the water drawdown inside the excavation should be taken into account because different codes and standards would allow different assumptions. The same argument holds for wall friction because individual national standards would have different requirements. The choice of the dilatancy angle,

Figure 1. Geometry of benchmark example Table 1. Material properties for soil and wall Soil Wall E kN/m2 30 000 3.0E7 - 0.3 27.5 0.18 - c kN/m2 kN/m3 10 20 / 19 - 24

Submitted results
Notes on individual submissions 13 results have been submitted, but as mentioned previously not all of them have provided the results in the form as specified. Because all entries are considered in the comparison presented in this paper a very brief summary of individual assumptions - as far as described by the authors - is given in the following. This should help explaining - at least to some extent - the differences observed in the results. No.1: gives no information at all on analysis. No.2: provides results for EC7 design approaches DA2 and DA3, uses different wall friction on active and passive side (2/3 and 1/3). Partial factor of 1.4 on soil weight on active side (does not conform to EC7 > possibly national annex (Italy)). No.3: does not apply any safety factors, wall friction 0.9. No.4: provides finite element and limit equilibrium results, global safety factor applied,

Plaxis Benchmark

Results from the ERTC7 benchmark exercise

Continuation

stiffness of strut/per unit length of excavation obviously different from specification, wall friction 0.8. No.5: compares finite element and beam spring results, loads, soil weight and strength parameters factored (similar to DA3) but with different factors than in EC7, cohesion and friction angle have different partial factors, wall friction 0.3, refers to Russian code of practice. No.6: applies Dutch CUR-method, which has some similarity with design approaches in EC7 but with different partial factors, wall friction 2/3. No.7: load factors on service loads according to British (1.4) or Australian standards (1.5), partial factor on soil strength of 1.2, provides additional results with subgrade reaction method, wall friction 2/3. No.8: provides finite difference and subgrade reaction analyses under various assumptions (e.g. undrained conditions), wall friction 2/3 and 1/3, does not explicitly provide design results, no safety factor given or applied to actions or parameters. No.9: a second submission applying the Dutch CUR-method, wall friction 2/3. No.10: compares EC7 design approaches DA2 and DA3 with partial factors according to EC7 and EAB (German recommendation for deep excavations) respectively, wall friction 0.5, assumes stiffness increasing with depth. No.11: does not apply a particular code but determines embedment from finite element analysis employing the strength reduction technique on the basis of a global factor of safety of 1.5, design bending moment obtained from calculated value at final excavation step multiplied by 1.5, wall friction 0.5. No.12: does not apply a particular code but determines embedment from finite element analysis employing the strength reduction technique on the basis of a global factor of safety of 2.0, wall friction 2/3, Drucker-Prager failure criterion for soil. No. 13: assumes in an alternative calculation capillary cohesion above groundwater level (determined from seepage analysis), design approach DA3 according to EC7 and analysis with characteristic soil parameters, wall friction 2/3. The lowering of the groundwater table was taken into account by means of phreatic levels by No.3, 5, 9, 10 and No.6, 7, 11 performed an additional interpolation in order to achieve continuous pore pressures at the base of the wall. No.4, 8, 12 and 13 performed a seepage analysis.

Results
Figure 2 shows the variation in embedment depth of all entries. As indicated in the previous section some submissions provided more than one solution (depending on assumptions made) and this is why there is more than one result plotted in these cases. It follows that the minimum embedment depth is 1.8 m and the maximum 5.5 m, but there is a tendency towards an embedment depth between 3.0 to 4.0 m.

Figure 2. Embedment length - all entries The maximum embedment depth of No.4 (5.5 m) is obtained by conventional analysis applying a global factor of safety = 1.5. The same embedment depth has been used in the numerical analysis. No.3 indicated a factor of safety of 1.76, No.2 factored the soil weight and No.12 obtained a factor of safety from strength reduction technique of 2.0. Thus these higher embedment depths can be explained by higher safety levels as compared to the other submissions. At the lower end (embedment depths around 2 m) it is found that these results are obtained from numerical analyses, either by means of a strength reduction technique or by factoring soil strength parameters. However, No.7 employed a factor of safety of = 1.2 which is lower than in EC7. No.13 indicated that the low depth is possible only by modelling the groundwater lowering by means of a 2-phase flow formulation resulting in suction effects above groundwater. From comments provided it can be concluded these two authors would not have designed a wall for a real project with this low embedment depth. Only 3 authors explicitly used EC7 for their analysis (DA2 and DA3), 3 authors used an approach similar to EC7 but with different partial factors of safety. It is somewhat encouraging to see that except for No.2 (where the soil weight was factored) the embedment depth is between 3.0 and 4.0m despite differences in assumptions of partial factors and wall friction. No.10 investigated the difference between DA2 and DA3 and obtained a slightly shorter wall with DA3.

Plaxis Benchmark

Conclusion
The results from a benchmark exercise addressing the design of a diaphragm wall for a deep excavation problem have been presented. Unfortunately not all of the submissions provided their results in form of design values but discussed the influence of various modelling assumptions without a clear statement on what they would use for design. However the results gave some valuable insight into the way design calculations are performed in different countries applying their respective codes of practice and standards and the differences in results are not surprising and not larger than in similar exercises which have been performed using conventional limit equilibrium analysis (Orr, 2005). It can be considered encouraging that the calculated embedment depth based on numerical analysis applying EC7 design approaches are within reasonable limits. Differences in bending moments and strut forces are less encouraging and clearly show the need for recommendations of good practice in numerical modelling in geotechnical engineering. Figure 3. Bending moments - all entries

References
Bauduin, C., De Vos, M. & Simpson, B. 2000. Some considerations on the use of finite element methods in ultimate limit state design. Proc. Int. Workshop on Limit State Design in Geotechnical Engineering, Melbourne. Bauduin, C., De Vos, M. & Frank, R. (2003). ULS and SLS design of embedded walls according to Eurocode 7. Proc.XIII ECSMGE, Prague (Czech Republic), Vol. 2, 41-46 Orr, T.L.L. 2005. Review of workshop on the evaluation of Eurocode 7. Proc. Int. Workshop on the Evaluation of Eurocode 7, Dublin, 31.3.-1.4.2005, 1-10. Figure 4. Strut forces - all entries Figure 3 shows bending moments and there we find a difference of approx. 300% but this includes values which are obviously not ULS-design values. This is however not surprising considering the different assumptions made in the various analyses. If we consider again the EC7 and related approaches (i.e. No. 5, 6, 9, 10 and 13) we see a range from 201 to 278 kNm/m for DA3. DA2 results in approximately 20% higher bending moments. If we multiply obvious SLS-bending moments by a partial factor of 1.35 we find that results also fall in this range. A similar picture is obtained for strut forces. Considering all solutions submitted the range is from 60 to 363 kN/m (including values representing SLS) and in this case also the EC7 and related approaches show a significant difference with the minimum being 194 and the maximum force being 363 kN/m. Schweiger, H.F. 2005. Application of FEM to ULS design (Eurocodes) in surface and near surface geotechnical structures. Proc. 11th Int. Conference of IACMAG, Turin, Italy, 19-24 June 2005. Bologna: Patron Editore. 419-430. Schweiger, H. F. 1998. Results from two geotechnical benchmark problems. Proc. 4th European Conf. Numerical Methods in Geotechnical Engineering, Cividini, A. (ed.), Springer, 645-654. Schweiger, H.F. 2002. Results from numerical benchmark exercises in geotechnics. Proc. 5th European Conf. Numerical Methods in Geotechnical Engineering (P. Mestat, ed.), Presses Ponts et chaussees, Paris, 2002, 305-314. Schweiger, H.F. 2006. Results from the ERTC7 benchmark exercises. Proc. NUMGE 2006.

Plaxis Practice

Modeling passive earth pressures on bridge abutments for nonlinear Seismic Soil Structure interaction using Plaxis
Anoosh Shamsabadi, Senior Bridge Engineer, Office of Earthquake Engineering, Sacramento, USA Steinar Nordal, Professor of civil engineering, Norwegian University of Science and Technology, Trondheim, Norway

Abstract
Seismic design of bridge structures is based on a displacement-performance philosophy. Full-scaled field experiments, observations from the major seismic events and post-earthquake studies have indicated that the bridge abutment-backfill exhibits strength and stiffness degradation. Hence it is important to incorporate the degradation of the abutment-backfill in the bridge global model in relation to its the seismic response and performance. This articles focus is on the development of stiffness degradation (nonlinear load deformation curve) for the earth pressure behind bridge abutments. This problem has recently been studied experimentally at several universities and research institutes in the USA and good-quality, large-scale test results are available. Backcalculations of the test results using Plaxis 3D Tunnel and 3D Foundations are shown herein. The Hardening Soil Model has been used to simulate the nonlinear abutmentbackfill force-deformation relationship from some of these tests. The Plaxis results are in good agreement with the experimental results and the simulations provide valuable information on the soil behavior and the soil parameters. It is concluded that Plaxis is an extremely valuable tool for the practicing bridge engineer in capturing realistic nonlinear soil-abutment-structure interaction curves.

Introduction
During a seismic event, the bridge deck moves laterally and collides with the abutment backwall. The passive resistance provided by the abutment reduces the displacement demands on the bridge columns. The basic idea of the displacement-performance-based design is to prevent total collapse by allowing selected components of the bridge system to yield with predictable large plastic deformations. Easily reparable components are preferably selected to yield. This concept allows the bridge engineers to limit the superstructure forces and displacements to acceptable levels. Often the abutment-backwall is designed to break off during a major seismic event in order to protect the abutment pile-foundation from plastic action. This implies that the passive earth pressure of a soil wedge in between the wingwalls is of special importance.

The passive earth pressure stiffness has been studied in experiments where external loads are applied by hydraulic jacks to push the abutment against the soil. The passive earth pressure stiffness has also been studied through analyzing strong motion records from bridges exposed to earthquakes. Seismic instruments have been installed in many bridge structures through out the world to improve our understanding of the global behavior and potential damage of the bridge structures during a seismic event. The information provided by monitoring structural responses led to better scientific understanding of nonlinear behavior of various components of the bridge system. Post-earthquake studies of various components of the bridge structures using system identification techniques has indicated that stiffness of the bridge abutment-backfill depends on the level of shaking and it varies significantly during the shaking. All the results indicate that significant nonlinearity and stiffness degradation occur within the bridge abutmentbackfill system due to hysteretic soil behavior. Therefore, reliable soil models are needed to predict the nonlinear force-deformation of the mobilized passive soil wedge as a function of bridge superstructure displacement in order to capture the global seismic behavior of the bridge structure model. Direct three-dimensional continuum finite element modeling of the surrounding soil and the bridge structure is by no means simple and computationally very expensive. As an alternative to the three-dimensional FEM, some simplifications are needed to enable the bridge engineer to simulate the global seismic response of the bridge structure. Beam elements are typically used to model columns and the superstructure. Simple and practical soil springs connected to the beam elements are widely used to represent the pile foundations. Therefore, the nonlinear abutment-backfill behavior should be modeled as nonlinear discrete soil springs as well. A nonlinear-continuum finite-element model can be used to develop the abutment-backfill nonlinear discrete soil springs in terms of backbone curves. In the global analysis of the bridge system the nonlinear hysteretic damping of the soil is then included using a Masing rule based on the nonlinear curves obtained from the soil continuum model. Figure 1 shows how the nonlinear discrete abutment springs are attached to a typical

Figure 1. Two-span box-girder bridge used in global bridge analysis

Plaxis Practice

Heaved Roadway and Backfill due to Passive Wedge

Figure 2. Seat-type abutment and foundation system nonskewed two-span single-column highway overcrossing bridge. Figure 2 shows a typical seat type bridge abutment, indicating how the bridge may move in the longitudinal direction and collide with the abutment backwall during a seismic event. The back wall is designed to shear off and allows the mobilized passive soil pressure to be developed as a result of backwall horizontal displacement. The effect of an actual earthquake is shown in Figure 3. This is an example of the passive wedge formed when a bridge superstructure is pushed into the abutment-backfill due to longitudinal seismic excitation. The surface cracks were developed in the roadway pavement behind the northern (23.5-m wide, near-normal 5o skew) abutment of the Shiwei Bridge in Taiwan during the Chi-Chi earthquake (Kosa et al, 2001). A typical highway bridge is wide and has a moderate back wall height, often 1 to 2 meters. The earth pressure problem is then a plane strain problem and 2D FEM simulations may be sufficient to simulate the soil response. However, for the skewed bridges the superstructure undergoes significant rotations about the vertical axis during seismic events that result in permanently lateral offset of the deck at seat-type abutments. Due to deck rotation the obtuse corners of the deck collides with the abutment wall and the acute corners of the deck move away from the abutment backwall. As a result asymmetric soil reactions are developed between the acute and obtuse corner of the abutment wall (Shamsabadi et. al, 2006), therefore, 3D analysis is needed to capture the nonlinear response of the backfill in between the wing walls for bridges with skewed abutments.

Figure 3. Passive wedges caused by an earthquake, Shiwei Bridge A series of full-scale static load tests was performed by Rollins and Cole (2006) on a 5.18 m by 3.05 m pile cap with a height of 1.12 m embedded in five types of soil backfills. The backfill was placed in layers and compacted to approximately 95% Modified Proctor density. The pile cap rested on a 3 by 4 group of 324-mm diameter steel pipe piles driven in saturated low-plasticity silts and clays. The resistance of the piles was subtracted from the measured total horizontal resistance of the pile cap to determine the mobilized passive resistance of the backfill. Depending on the backfill properties, the measured earth pressure force on the back wall varied from 750 kN to 2000 kN at a horizontal wall displacement of about 6 cm. Figure 4 shows the schematics of a typical BYU pile cap field experiment with the threedimensional passive wedge bounded by a logarithmic-spiral type failure surface, based on field measurements of observed cracking patterns and wedge deformations by Rollins and Cole (2006).

Mobilized limit-equilibrium methods

The results of the experimental nonlinear force-deformation response from a number of full-scale tests, centrifuge model tests and small-scale laboratory tests for walls, bridge abutments and pile caps in a variety of structure backfills have been studied using a model by Shamsabadi et al. (2006). This two-dimensional plane-strain model is based on limit-equilibrium using a Logarithmic-Spiral failure surface coupled with a modified Hyperbolic stress-strain relationship (LSH). Good agreement was found between the LSH model and the experimental data. The computed results from the LSH model were multiplied by an adjustment factor a varying between 1.2 and 1.4 to account for the three-dimensional shape of the mobilized passive wedge in the backfill (see Figure 4). The force-deformation relationships calculated using 3D Plaxis were compared against experimental data without any adjustment factor.

Plaxis simulations of experimental data

Passive resistance behind walls has been studied in centrifuge tests (such as by Gadre and Dobry, 1998), laboratory tests (such as by Fang et al, 1994), and large-scale or fullscale tests (such as by Romstad et al., 1995 and documented by Martin et al., 1997). Recent full-scale tests at Bingham Young University (BYU) are reported by Rollins and Cole (2006). Large-scale tests are currently being performed at University of California San Diego and at University of California Los Angeles to capture the nonlinear forcedeformation characteristics of the abutment-backfill for the seismic design of bridge structures.

Plaxis Practice

Modeling passive earth pressures on bridge abutments for nonlinear Seismic Soil Structure interaction using Plaxis
Continuation

Computation of stiffness parameters

In the Hardening Soil Model, the applied stiffness for sand follows expressions of this type: E50 = E50ref 3 + a (1) pref + a Where a = c / tan , 3 is the minor principal stress, and the reference pressure is pref = 100 kPa. The applied stiffness in the Hardening Soil model in Plaxis is hence not only a function of the input stiffness parameter, E50ref, but also a function of the stress level and the cohesion, c. The reference input stiffness parameters E50ref for sand and gravel are normally in the range 15 MPa to 50 MPa. Here significantly higher values are used, actually 100 MPa, except for the silty sand. The backfill material is well compacted, which justifies high stiffness. However, the main reason for selecting higher stiffness values is related to the low initial stresses behind the wall. When the term 3+ a in equation 1 is low then the applied stiffness, E50, will be low. Conventional experience on using reference stiffness in the range 15 to 50 MPa is primarily obtained from natural sands and valid primarily at higher stress levels, (3- values). The present experimental results indicate that for problems with low 3- stresses, higher input stiffness parameters than normal may be relevant. Some further comments are given in the following paragraphs regarding the selection of stiffness parameters. Observation of the BYU nonlinear experimental force-deformation relationship indicates an average stiffness (k50) of about 100 kN/mm for the sand and gravel backfill. Assume that the deformation behind a wall with the height H is equal to dh due to an average horizontal strain within an influence zone L. Let L be in the order of 2H. An average stiffness (E) may then be introduced by the following equation: dh = h E P Pp 2Pp 2k50 h L = h 2H = p 2H = 2H = = E AE WHE WE WE (2)

Figure 4. Full-Scale Static BYU Load Test on Silty-Sand Backfill

Input parameters to finite element studies

The pilecap experiments conducted at (BYU) demonstrated significant 3D effects as shown in Figure 5 . Both the Plaxis 3D Tunnel and Plaxis 3D Foundation were used to simulate the pile cap experiments. Table 1 shows the backfill strength values and the stiffness parameters used in the Hardening Soil model available in Plaxis. Selection of the stiffness parameters used in Plaxis is a compromise between simplicity and a reasonable match with experimental data. Perfect match could have been obtained if strength and stiffness variation were utilized for every separate test. But keeping the strength the same as proposed by the experimentalists and using the stiffness parameters shown in table 1, the results obtained using 3D Plaxis are in good agreement with experimental data. The stiffness parameters listed in table 1 provide realistic force-deformation relationships of the bridge abutment-backfill and appears reasonable for practical design. For the parameters not listed in table 1 the default values recommended in the Plaxis user manuals are used. The plate element with linear elastic and high stiffness is used to model the rigid back wall.

Table 1. Input parameters for 2D and 3D Plaxis analyses BYU Soil g c Rinter Rf E50ref [MPa] Eurref [MPa]

[kN/m3] Friction [kPa] Dilatancy Wall interface Clean Sand 18.4 390 4 90 0.70 0 Fine Gravel 20.8 34 4 40 0.70 Coarse Gravel 23.2 400 12 100 0.70 Silty Sand 19.2 270 31 00 0.70

0.97 0.97 0.97 0.97

100 200 100 200 100 200 50 100

10

Plaxis Practice

where Pp is 50% of the ultimate passive earth pressure force acting over a total back wall area of A = WH, and h = Pp /A. The fraction 50% is chosen to give a half way to failure average stiffness. The wall width W is 5,180 mm in the BYU test (Figure 4). Turning equation 2 around we obtain: E= 2k50 W = 2.100kN / mm= 40MPa 5180mm (3)

Computed results
Both the 3D Tunnel and the 3D Foundation programs were used in this study. The simulations were done using a half-wall-width model due to symmetry of the problem. Figure 6 shows the half-model of the geometry used with a medium coarse mesh as applied in the 3D Foundation program. The use of a finer mesh did not change the results significantly, but made convergence at large deformations considerably more difficult.

Sequence of the events

All the analysis was performed in steps to simulate sequence of the real events during the field experiments. The computations were simulated using the following steps: 1- Starting with a level ground the initial stresses were calculated. 2- Excavation was performed by deactivating clusters of elements in the front of the pile cap. 3- In 3D Tunnel, displacements were applied while in 3D Foundations distributed normal stresses were applied to push the pile cap up to backfill failure. Before the field experiment, the backfill was placed and compacted behind the wall. The backfill was extended to the sides behind the pilecap, but no backfill was placed along the sides of the pile cap. This is why the simulations were performed as shown in Figure 5. The free vertical soil face in the model is supported by a Ko pressure and increased linearly below ground surface. This pressure is kept constant throughout the test. The wall was pushed then horizontally without any vertical movement. In Plaxis 3D Foundations, horizontal translation was obtained by adding an extended handle normal to the wall plate (see Figure 5) and specifying the appropriate horizontal line fixity on the handles lower edge. The simulated load deformation curves are shown in Figure 6 to Figure 9. Figure 6 shows the sandy gravel experiment, giving an ultimate load for the 5.18 x 1.12 m2 wall of about 1000 kN. The initial stiffness is in the order of 1500 kN/cm which corresponds to about 250 kPa/cm. Figure 7 shows results for Coarse Grained Gravel and Figure 8 shows results for Fine Grained Gravel where the ultimate capacity is about 2000 kN and 780 KN, respectively. The initial stiffness is however about the same as for Sandy Gravel. Figure 9 shows the result of the simulation for the silty sand backfill. There is a good agreement between the Plaxis simulations and the experimental curves. Some of the load-deformation curves simulated by both Plaxis 3D Foundation and Tunnel are not smooth and show some minor irregularities. Some of this is related to the development of alternative plastic failure mechanisms as illustrated by Figure 10 and Figure 11. Figure 10 shows two failure competing mechanisms. Figure 11 shows that one mechanism becomes more dominant as deformation increases. Figure 7 shows the change in the inclination of the simulated nonlinear load-deformation curve at about 3 cm. This is a reflection of the fluctuating failure mechanism. Such behavior was seen in several simulations and is believed to be attributed to the inherent minor instability of non-associative plasticity. The authors believe that such feature may also be seen in real soil behavior.

We note that 40MPa is a stiffness considerably lower than the input values used, but the average stiffness of 40MPa is not a reference stiffness. Applying rough, average parameters (a=5 kPa, pref=100 kPa, 3=15 kPa), the input stiffness parameter E50ref can be estimated to: E50ref = E50 pref + a h + a 100kPa + 5kPa = 40MPa = 90MPa 15kPa + 5kPa (4)

Please keep in mind that this is a crude estimate. It may still indicate why we apply the input reference stiffness (100 MPa ) for the high-quality gravel and sand backfill. The parameters used in the Hardening Soil model are identical to the parameters used in the LSH model proposed by Shamsabadi et al (2006). Following the recommendation by Shamsabadi et al. (2006), the Hyperbola cut off parameter, Rf, was set equal to 0.97. This gives a better match with the measured nonlinear force-deformation than the default value of 0.90 proposed in the Plaxis user manual. A pre-consolidation effect due to compaction of the backfill was originally used, but resulted in a rather sharp drop in stiffness as the applied load was increased beyond the pre-consolidation pressure. Such a drop is not observed in the experimental data. Therefore, the OCR was set equal to one and the effect of the backfill compaction was taken into account using the stiffness parameters listed in table 1.

Figure 5. A half model of the BYU field test, medium coarse mesh, 3D Foundations

11

Plaxis Practice

Modeling passive earth pressures on bridge abutments for nonlinear Seismic Soil Structure interaction using Plaxis
Continuation

Figure 12 show examples of the logarithmic spiral failure surface mechanism using the Plaxis 3D Tunnel. There is a good match between simulated deformed shapes of the passive wedges from the 3D Plaxis and the BYU pile cap experiments mapped in the field.

Test data 3D Tunnel 3D Found

Test data 3D Tunnel 3D Found

Figure 8. Load deflection curves for Fine Grained Gravel

Figure 6. Load deflection curves for Clean Sand Test data 3D Tunnel 3D Found

Figure 9. Load deflection curves for Silty Sand Test data 3D Tunnel 3D Found

Figure 7. Load deflection curves for Coarse Grained Gravel

Figure 10. Coarse Grained Gravel, incremental shear strain contours at 3 cm deformation 12

Plaxis Practice

(a) Typical deformed mesh with total displacement contours

Figure 11. Coarse Grained Gravel, incremental shear strain contours at 6 cm deformation

Simulation time
The computer simulations were fast using the Plaxis default coarse mesh. The coarse mesh was found to be too rough and mesh refinement with several mesh densities have been tried. The time required for each computer simulation depends primarily on the mesh density but also on the selection of parameters for the iterative procedure to obtain proper convergence. Convergence was more slow for certain combinations of stiffness, friction, cohesion and dilatancy parameters. Still coarse meshes were analyzed in a few minutes while the ones shown above take from half an hour to more than an hour on a Personal Computer. It is confirmed that simulating an indentation problem to large deformations are numerically challenging and hence manual iteration control was used. The overrelaxation factor was often set to 1.0 while low values were selected for the desired number of iterations. Some of the results are obtained by using a tolerated error of 3%. It was observed that using high tolerated errors even up to 10% did not change the computed results significantly compared to a 1% tolerated error analysis. A proposed explanation for this may be that divergence tendencies, when they occur, appears to be related to zones with tension and stress concentrations in corners, not so much to the volume of soil mainly providing the passive resistance. Still, high tolerated errors should never be used uncritically. (b) Typical shaded plot of total displacement contours

(c) Cross section of incremental shear strain

Mesh size
3D simulations accentuate mesh dependency since normally a rather low number of elements is used to avoid excessive computation times. At the same time 3D simulations require lower order elements. In order to get a feeling for the effect of mesh refinement and compare 2D simulations to 3D simulations, a true 2D problem was analyzed using both Plaxis 2D and Plaxis 3D Foundations, Figure 13 and Figure 14. Three different mesh densities were used in 3D with a Mohr Coulomb soil with = 22 KN/m3, E = 30 MPa, = 0.3, c = 20 kPa, = 38,70 and = 100. Drained conditions and no groundwater are assumed. The same soil parameters and a very fine mesh were used for Plaxis 2D.

(d) Cross section of plastic points in the incremental shear strain zone Figure 12. Results from a half wall width simulation with 3D Tunnel.

13

Plaxis Practice

Modeling passive earth pressures on bridge abutments for nonlinear Seismic Soil Structure interaction using Plaxis
Continuation

All the load deflection curves given in Figure 14 are identical up until about 50% of the capacity where after they deviate. The deviation is considerable for the coarse 3D mesh. This might be expected since it is very coarse with only two elements over the wall height. Some uncertainty is related to the irregularity of the curves. The medium 3D mesh appears to perform satisfactorily, and the overshoot for rather crude meshes is less than 20%. Some overshoot can be expected unless very fine meshes are used. If the expected percentage of overshoot is roughly known, a reasonably good simulation can be achieved using a medium dense mesh. For geotechnical applications, a medium dense mesh appears to be an decent trade-off between accuracy and computing time.

Conclusions
A full 3D Plaxis stiffness degradation model for the bridge abutment-backfill has been presented. The model simulates stiffness degradation behavior of the abutment-backfill to replicate the stiffness the degradation which has been observed during a major seismic event. The authors found that there is a very good agreement between the nonlinear experimental force-deformation behavior of the mobilized passive wedge and the 3D Plaxis simulations using the Hardening Soil Model. For simulation of wide bridge abutments where plane-strain conditions can be assumed, it is the authors recommendation that the Plaxis 2D should be used without any adjustment factors for the 3D effects. In other cases such as skewed abutment due to bridge rotation, a 3D simulation is more relevant. Based on the full-scaled experiments of the more narrow abutments studied in this article, the 3D effect can result in increase of 20 to 40% in passive soil resistance compared to a plane-strain case. The analyses indicate that 3D simulations give realistic results. A medium mesh size is found to give reasonable results for geotechnical applications. The Plaxis results are in good agreement with the experimental results and the simulations provide valuable information on the soil parameters and the failure mechanism. Plaxis can be a valuable analysis tool for the researchers and practicing bridge engineers to evaluate realistic nonlinear soilabutment-structure interaction behavior. (a) Coarse mesh

(b) Medium mesh

{c) Fine mesh Figure 13. Effect of mesh sensitivity using Plaxis 3D Foundations

Figure 14. Load deflection curves illustrating mesh dependency

14

Plaxis Practice

References:
PLAXIS 2D PLAXIS 3D Tunnel PLAXIS 3D Foundations Fang, Y.-S., Chen, T.-J., and Wu, B.-F. (1994). Passive Earth Pressures with Various Wall Movements. Journal of Geotechnical Engineering, ASCE, Vol. 120, No. 8, pp. 1307-1323. Gadre, A. D., and Dobry, R. (1998). Centrifuge Modeling of Cyclic Lateral Response of PileCap Systems and Seat-Type Abutments in Dry Sand. Report MCEER-98-0010, Rensaeler Institute, Civil Engineering Dept., Troy, NY. Gadre, A.D. (1997), Lateral Response of Pilecap Foundation System and Seat-type Abutment in Dry Sand, PhD Dissertation, RPI. Martin, G. R., Yan, L.-P., and Lam, I. P. (1997). Development and Implementation of Improved Seismic Design and Retrofit Procedures for Bridge Abutments, Final Report, Research Project Funded by the California Department of Transportation, September. Romstad, K., Kutter, B., Maroney, B., Vanderbilt, E., Griggs, M., and Chai, Y. H. (1995). Experimental Measurements of Bridge Abutment Behavior. Department of Civil and Environmental Engineering, University of California, Davis, Report No. UCD-STR-95-1, September. Rollins, K.M., and Cole, R.T. (2006), Cyclic Lateral Load Behavior of a Pile Cap and Backfill, accepted for publication, Journal of Geotechnical and Geoenvironmental Engineering. Shamsabadi, A., Kapuskar, M. and A. Zand (2006), Three-dimensional Nonlinear Finite Element Soil-Abutment-Structure Interaction for SkewedBridges, Fifth National Seismic Conference on Bridges & Highways September 18-20, 2006 in San Francisco Shamsabadi, A., Rollins, K., and Kapuskar, M. (2006), Nonlinear Soil-Abutment-Bridge Structure Interaction for Seismic Performance-Based Design, Journal of Geotechnical and Geoenvironmental Engineering, ASCE (under review for publication). Shamsabadi, A., and Kapuskar, M. (2006), Practical Nonlinear Abutment Model for Seismic Soil-Structure Interaction Analysis, 4th Intl Workshop on Seismic Design and Retrofit of Transportation Facilities, San Francisco, March 13-14.

15

Plaxis Practice

Prediction of soil deformations during excavation works for the renovation of Het Nieuwe Rijksmuseum in Amsterdam, The Netherlands
Dipl.-Ing. H.D. Netzel (CRUX Engineering b.v.) Ir. D. Vink (CRUX Engineering b.v.)

Introduction
The Rijksmuseum in Amsterdam is one of the most important 19th century monumental buildings in The Netherlands. The showpiece of the museum is the world famous The Night Watch by Rembrandt. Presently the building is being renovated in order to meet the modern standards for international museums. The design of the renovation has been drawn by Cruz y Ortiz from Sevilla. The constructive design is being carried out by ARCADIS in cooperation with CRUX Engineering and WARECO for the assessment of the geotechnical and hydrological implications of the project. The most striking feature in the new design is a semi-underground square of approximately 3000 m2 in between the existing monumental building. This square is situated in the existing court yards and will serve as main entrance to the museum. The geotechnical design calculations are carried out by CRUX using the Finite Element program PLAXIS. The calculations are part of the risk assessment strategy in order to predict and judge the influence of ground deformations due to the excavations on the surrounding building. Intensive monitoring is used to control the deformations of the structure and the sheet pile walls during the construction work.

Figure 2. Cross section the near vicinity of the pits. These deformations, caused by the excavation works, have been predicted using PLAXIS. Figure 1 shows an artist impression of the court yards with lowered ground level. Figure 2 shows the museum in cross section. Below the court yards cellars will be realised for use as conference rooms, an auditorium and the kitchen.

Soil Conditions
For the project 18 electric cone penetration tests (CPT) have been performed. The CPTs show the soil profile that is typical for Amsterdam. The top layer of 2 m to 3 m below surface level consists of Anthropogenic sand. Below this toplayer the Holocene deposits are found to a depth of about 13 m below surface level. The Holocene formation can be divided (from top to bottom) into 1 m to 2 m peat (Hollandveen), 1 m to 2 m clay (Oude Zeeklei), 3 m silty sand (Wadzand), 3 m clay (Hydrobiaklei) and 0,4 m peat (Basisveen). The soft Holocene has been deposited on top of the stiff Pleistocene sands consisting of the so called first sand layer of 2 to 3 m thickness, an intermediate silty, clayey sand layer (Allerod) of 2 m and the second sand layer of at least 5 m thickness. The phreatic water level is about 0,4 m below surface level. The artesian water level in the first sand layer lies 2 m below the phreatic water level.

General
The realization of the lowered square and underground facility rooms requires the construction of concrete cellars with a depth of 6 m or more, right next to the existing wooden foundation piles of the monumental building. The cellars will be cast in building pits, the excavation of which will cause relaxation of the soil, inducing deformations in

PLAXIS Model
A schematic soil profile that best fits the CPTs has been constructed for the model geometry. For modelling the soil layers the Hardening Soil model has been adopted. This model is most suitable for excavations. It is capable of describing reduction of stiffness as well as irreversible deviator strains due to deviatoric stress. In the vicinity of the excavation (on the active side) deviatoric stress applies. The horizontal stresses are reduced more compared to the vertical stresses. This results in friction hardening which is a feature of the Hardening Soil model. Friction hardening also occurs at the passive side of the sheet pile (that is, below the bottom of the pit). Here horizontal stress increases compared to the vertical stress due to (horizontal) movement of the sheet pile on the one hand and reduction of vertical stress as a result of the excavation on the other hand. In order to get parameters for the Hardening Soil material set, almost 50 samples have been taken from 13 borings. Four soil layers have been tested: Hollandveen, Oude Zeeklei, Wadzand and Hydrobiaklei. These layers are considered to be most important regarding deformations due to the excavation. Besides testing the volumetric weight (), triaxial tests and oedometer tests have been performed for determining stiffness parameters (E50, Eoed) and strength parameters (c, ). The parameters from the laboratory tests

Figure 1. Artist impression of the lowered court yards (source JNDG-Amsterdam)

16

Plaxis Practice

The cross section shown in Figure 2 has been modelled in Plaxis 2D using the plane strain model consisting of a mesh of triangular elements with 15 nodes. In total 89 clusters have been defined. The generated mesh contains1906 elements. A picture of the model geometry is shown in Figure 4. Due to the considerable length of the cross section only part of it has been modelled in order to reduce the number of elements and calculation time. The western court yard with its sloping bottom has been entirely included in the model geometry; however only one symmetric half of the eastern court yard has been taken into account. The distance between the boundaries on the left and the right side of the model geometry is 100 m. The bottom level is set at NAP 30 m; the top level at NAP +0,0 m. No existing structural elements such as foundation piles, foundation beams or floors have been modelled. Hence, the model considers a green field calculation in which the stiffness of these elements is neglected. This is a conservative approach regarding the expected differential deformations of the adjacent structure. Existing loads from the building on the stiff sands at pile tip level have however been taken into account. The sheet pile walls (three in number) have been modelled as a beam element with the characteristics of AZ 18 profiles. The piles of the underwater concrete floor are loaded by tensional forces during pumping dry the building pits. Anchor elements and geotextile

Figure 3. Characteristic Soil profile have been compared to available parameter sets from other projects in the central part of Amsterdam. They generally show a good agreement. For the calculations the mean values of stiffness have been taken whereas for the strength the low representative values apply. Table 1 contains the soil parameters for the Hardening Soil material data set. Table 1. Hardening Soil parameter set

Type unset set kx ky E50ref Eoedref Eurref cref 3 3 3 2 2 2 2 [kN/m ] [kN/m ] [kN/m ] [m/day] [m/day] [kN/m ] [kN/m ] [kN/m ] [kN/m ] [] [] 1 1st sand layer Drained 19,8 19,8 1,3E+01 1,3E+01 35000 20000 100000 1,0 33,0 3 2 2nd sand layer Drained 19 19 8,0E+00 8,0E+00 32000 25000 80000 1,0 33,0 3 3 Allerod Undrained 18,5 18,5 2,6E+00 2,6E+00 15000 9140 30000 3,0 28,0 0 4 Hollandveen Undrained 10,5 10,5 1,7E-03 8,6E-04 2200 1187 8817 12,1 18,2 0 5 Top layer Drained 17 18,4 8,6E-01 8,6E-01 17000 15000 50000 1,0 27,0 0 6 oude zeeklei Undrained 15,4 15,4 1,3E-04 1,3E-04 9400 5429 20000 2,8 27,4 0 7 wadzand Undrained 18,1 18,1 8,6E-03 8,6E-03 11600 4904 25000 2,0 34,9 4,9 8 Hydrobiaklei Undrained 14,7 14,7 9,0E-05 9,0E-05 5700 3791 11400 13,3 23,3 0 11 basisveen Undrained 11,7 11,7 1,0E-03 1,0E-03 2000 1004 7000 6,0 21,0 0 ID

Name

ID Name 1 1st sand layer 2 2nd sand layer 3 Allerod 4 Hollandveen 5 ophooglaag 6 oude zeeklei 7 wadzand 8 Hydrobiaklei 11 basisveen

kur [-] 0,2 0,2 0,2 0,15 0,15 0,15 0,2 0,15 0,2

pref [kN/m2] 100 100 100 100 100 100 100 100 100

Power [-] 0,5 0,5 0,5 0,72 0,8 0,57 0,75 0,59 0,8

Ko:nc [-] 0,40 0,46 0,50 0,69 0,55 0,54 0,43 0,60 0,65

c_incr [kN/m3] 0 0 0 0 0 0 0 0 0

yref [m] 0 0 0 0 0 0 0 0 0

Pf [-] 0,9 0,9 0,9 0,9 0,9 0,9 0,9 0,9 0,9

T-Strength [kN/m2] 0 0 0 0 0 0 0 0 0

Rinter [-] 0,67 0,375 0,67 0,5 0,67 0,67 0,67 0,67 0,5

17

Plaxis Practice

Prediction of soil deformations during excavation works for the renovation of Het Nieuwe Rijksmuseum in Amsterdam, The Netherlands
Continuation

Risk assessment of influence on the existing building and its foundation

The calculated greenfield deformations are used to access the influence on the existing Rijksmuseum building. The following aspects are considered: - Horizontal soil displacements can cause lateral pile loading affecting the bending capacity of the existing wooden foundation piles of the Rijksmuseum. - Vertical soil displacements on surface and pile toe level (1st sand layer) can cause pile and consequently (differential) building deformations. Vertical soil displacement at surface level needs to be controlled because settling soil may lead to an increase in negative shaft friction on the existing wooden foundation piles. - Reduction of effective stress at pile tip level due to the excavation may lead to reduced bearing capacity of the end-bearing piles of the Rijksmuseum close to the excavation. These aspects are analysed with numerical, analytical and empirical prediction methods in order to quantify the damage risks.

Figure 4. Model geometry elements have been chosen to model the tension piles. The tensional stiffness of these elements has been adapted as to match the stiffness of the line of piles by dividing the axial pile stiffness by the center-to-center distance of 2,5 m of the piles.

Construction phases
Table 2 shows the construction phases with a short description. Except for the consolidation phases, all phases are of plastic analysis type. Consolidation phases have been defined in order to allow for dissipation of excess pore (under)pressures that develop during excavation.

Results
For all main construction stages the following relevant deformations and stresses have been read from the output: - horizontal deformation of the sheet pile wall; - horizontal and vertical deformations in a horizontal cross-section at surface level; - vertical deformation in a horizontal cross-section at pile tip level;

Table 2: Overview of construction phases Phase foundation load on and load passage on load passage off sheet pile walls on East: strut +0,0; excavation 2,6; water level -3,8 West: strut +0,5; excavation 2,6; water level -3,8 consolidation 21d water level -0,8 Description Soil behaviour loads from pile groups and load of passage activated (historic higher surface level) load of passage deactivated in accordance with actual surface level beams activated; reset displacements from previous stages to zero strut court yard east activated; lowering of water level in eastern pit to NAP3,8 m and excavation to NAP2,6m (this dry excavation step is necessary due to the soil decontamination in the above layers) strut court yard west activated; lowering of water level in western pit to NAP3,8 m and excavation to NAP2,6m (this dry excavation step is necessary due to the soil decontamination in the above layers) consolidation phase, 21 days water level in both pits rises from NAP3,8m to NAP0,8m drained drained drained

undrained

undrained consolidation undrained undrained undrained consolidation undrained undrained undrained

East: excavation -7,0; water level 0,8 Final excavation pit east to NAP7,0m; water level in pit remains NAP0,8m West: excavation -7,7/-7,0; water level 0,8 Final excavation pit west to depth of NAP-7,0m to NAP7,7m; water level in pit remains NAP0,8m consolidation 30d East & West: underwater concrete East: pump dry West: pump dry consolidation phase, 30 days underwater concrete clusters activated and material set applied water in pit east is removed water in pit west is removed

18

Plaxis Practice

- modification of effective soil stresses near pile tips of the existing piles of the Rijksmuseum. Figure 5 shows the deformed model and the contourplot of the deformations in the fully excavated situation. Figure 6 shows the lateral deflections of the sheet pile wall in different stages of the excavation.

Figure 6. Lateral deformation of the sheet pile wall

Conclusion
In order to predict soil deformations in the vicinity of two excavation sites that are located inside the monumental building of the Rijksmuseum in Amsterdam extensive PLAXIS calculations have been performed. The design of the excavation of the building pits with an underwater concrete floor is predicted to cause an acceptable influence on the adjacent structure. The locations and the extent of the monitoring of the building and the sheet pile wall was derived from these risk analysis. The performance of the construction work on the adjacent building will be frequently compared with the predicted deformations in order to be able to anticipate in time with regard to unexpected deformation trends during the construction process.

Figure 5. shadings plot of total deformations in final construction phase

The calculated maximum deformations are summarized in Table 3. These greenfield results are translated to the pile foundations and the adjacent building of the Rijksmuseum in order to quantify the impact on the adjacent structrure. The maximum cumulative deformation due to the different sources is predicted to be 5mm and the angular distortion along the building is restricted to be 1/3000. The risk of damage at the masonry structures is negligible and the bending capacity of the wooden piles is sufficient in order to withstand the additional lateral pile loading due to the excavation.

References:
[1] A.M. De Roo, H.D. Netzel, P.J.M. Den Nijs; Omgaan met risicos bij de renovatie van Het Nieuwe Rijksmuseum; in Dutch; journal GEOTECHNIEK edition 3, juli 2006.

Acknowledgments:
The project organisation Het Nieuwe Rijkmuseum is greatfully acknowledged for providing the permittion to publish this article.

Table 3: maximum deformation Type and location Value Maximum horizontal soil displacement at surface level at the location of the piles of the existing building (at closest distance to the sheet pile wall) 17mm Maximum surface settlement at the location of the piles of the existing building (at closest distance to the sheet pile wall) 12mm Maximum settlement on pile toe level at the location of the piles of the existing building (at closest distance to the sheet pile wall) 2mm Maximum lateral wall deflection of the sheet pile wall 29mm Reduction of vertical effective soil stress at at the location of the pile toes of the existing building (at closest distance to the sheet pile wall) 15%

19

Plaxis Practice

Comparison of the effectiveness of Deep Soil Mix columns using 2-D and 3-D Plaxis
Hester Leung, Cristien Gani, Wataru Okada SKM Auckland Sergei Terzaghi, SKM Sydney

Introduction
The purpose of this paper is to compare the behaviour of deep soil mixing columns predicted by Plaxis 2D and 3D programmes in terms of stresses acting in columns and calculated factor of safety (FoS). Deep soil mixing column has been used as one of the remedial solutions on slip repair works of state highways in the Northland region of New Zealand. Slope failure is common in Northland region due to the unique characteristic of the geology in this area and also the heavy rainfall experienced particularly in winter months. Prior to installing the deep soil mixing columns, the road had to be repaired regularly by smoothing of the pavement. A heavy rain event eventually caused sufficient significant damage to the road such that slope stabilisation work was considered necessary. One such study was selected to both illustrate the repair method, integration with Plaxis, and also to illustrate the issues with the use of FE technology in routine design. Experience to date suggests that conventional limit equilibrium design was not capturing the field behaviour of the columns, and greater understanding of the real behaviour was needed. In this particular study, 2-D and 3-D Plaxis was used, as concerns were raised about the suitability of only two rows of columns in the design by the project reviewers. Field experience also indicates that the effectiveness of the columns is bounded. If columns are too close there is no further gain in overall strength, and too far apart, and the columns begin behaving individually with no net benefits. This paper illustrates some of these issues.

Figure 1. Slip on the shoulder of the road

Slip Descriptions
Topography and Ground Condition The slip discussed in this paper occurred in State Highway 11, Northland. It was a road slip which occurred in a section where the road has been constructed on a natural valley feature. The upper slope of the valley above the road is hummocky and shows a number of headscarp, creep and slump features. The area below the failure is highly vegetated with trees and bush, with the lower flanks of the valley also containing houses. Borehole logs indicate that the sub surface condition consist of soft to firm silty clay sandwiched between fine to coarse grained gravel and firm to stiff silty clay. Mudstone/sandstone forms the bedrock. Failure Mechanism Site observations suggest that the failure is likely to have resulted from a circular slip, which has probably been induced by saturation of the slope during a heavy rainfall event. Creep movement was noted below the toe of the slip, which would have contributed to the failure to some extent. A secondary deeper block failure was suspected to be responsible for the crack adjacent to the road centreline. The slip is approximately 50 m long in the longitudinal direction. Photographs of the slip are shown in Figures 1 and 2.

Figure 2. Close up of the slip Once the actual slope failure is successfully simulated, deep soil mixing columns are introduced into the model. The columns are modelled as soil using volume elements. Typically the design unconfined compressive strength of a deep soil mixing column is about 1.5 MPa. In reality, the achieved strength is much higher in most cases. In 2D models, the column has been modelled using a replacement ratio method in the out of plane direction. The columns are modelled together with the surrounding soil as a block of composite material. This takes the column spacing into account by assigning appropriate average composite properties. The 3D model uses the design properties of the deep soil mixing column as input. The columns are positioned in such a way that a block of improved soil bounded by the columns is created in the pavement and the slope below the road. For the purpose of this paper, the slope is assumed to be dry. Adjustments have been made to the subsurface parameters such that the dry slope models have a pre-existing failure mechanism similar to the interpreted mechanism when taking groundwater into consideration. Two rows of 0.3 m by 0.3 m columns are installed and the lengths of the columns are 5.0 m and 6.0 m. Our research focus on the behaviour of columns for the

Deep Soil Mixing Column Assessment

Typically, geotechnical parameters used in the design are derived from laboratory test results of selected core samples collected during the site investigation. In some cases, geotechnical parameters are assessed from other site information based on the geology and site observation. An initial analysis is carried out to verify the geotechnical parameters used by simulating the actual slip failure in the model.

20

Plaxis Practice

following scenarios: - Variation on column spacing i.e. 0.6 m (2 diameter spacing), 2.5 m (8 diameter spacing) and 4.0 m (13 diameter spacing) in the longitudinal direction (parallel to road). The 2.5 m spacing is the practical maximum spacing we have adopted based on field observations of completed projects. - The effect of soil model used in the assessment i.e. Mohr Coulomb Soil Model and Hardening Soil Model.

Table 2. Comparison of maximum shear acting in columns Column Spacing (m) Maximum Shear Stress (kPa) Mohr Coulomb Hardening Soil Model 2D 3D 2D 3D 0.6 142 70 111 64 2.5 326 73 252 54 4.0 388 70 287 50

Figure 3. Cross section and parameters used in the model

Figure 4. Failure Mechanism

Results
The output of Plaxis analysis for different scenarios with regard to factor of safety and shear stress acting in column are given in Tables 1 and 2 respectively. The factor of safety prior to columns installation is 1.1 and 1.2 as determined from 2D and 3D modelling respectively. Both models give a similar failure mechanism as shown in Figure 4. The shear stress mobilised by the columns are determined from the deviatoric stress given in the Plaxis output. In 2D models, this needs to be converted to an equivalent shear stress as the columns have been modelled as a block of composite material instead of an individual column as in the case for 3D. The conversion is based on the assumptions that the shear stress taken by each column is proportional to the stiffness of surrounding soil and takes into consideration of column spacing. This aspect needs to be refined to maximise the effectiveness of the columns. Table 1. Comparison of factor of safety using Plaxis 2D version 8.2 and Plaxis 3D Tunnel Programme Column Spacing (m) Factor of Safety (FoS) Mohr Coulomb Hardening Soil Model 2D 3D 2D 3D Pre-existing Case 1.10 1.21 1.10 1.19 0.6 1.54 1.55 1.53 1.54 2.5 1.53 1.45 1.52 1.45 4.0 1.54 1.43 1.51 1.44

Discussion
Factor of Safety Table 1 indicates that the computed factor of safety in 3D models are lower than 2D models by 5% to 7%, however this is not reflected for model with the column spaced at 0.6 m. This is because on 3D model, the actual position of column is modelled whilst on 2D model the variation of column spacing is modelled by adjusting the composite column parameters, thus in 2D model the effect of stress distribution for different column spacings have not been taken into account. Already we see the impact of the bounds noted above. For this problem a spacing of 2.5 m is the maximum, and probably a closer spacing would be more beneficial in reality. No significant difference is noted in terms of soil model used in the analysis. Both the Mohr Coulomb Model and Hardening Soil Model give in similar results for this particular slip. This is possibly because the problem is modelled as dry thus the strength reduction (and consequent change in stiffness due to strain) due to groundwater has not been seen in the model which otherwise will be better represented by Hardening Soil Model. Shear Stresses in Column 2D models show significantly higher shear stress compared to 3D model as shown in Table 2 above because the shear stress in 2D model is the equivalent shear stresses acting in columns which calculated using the equivalent stiffness assumption whereas the shear stress in 3D model is the actual stress determined from the finite element analysis. Mohr Coulomb models show higher shear stresses in the column in comparison to Hardening Soil Model, if E of the Mohr Coulomb model is equal to E50ref of the Hardening Soil Model. However, the impact of the columns being stiffer than the surrounding soil modifies the stress regime and hence the actual stiffness, making the direct comparison more difficult.

21

Plaxis Practice

Comparison of the effectiveness of Deep Soil Mix columns using 2-D and 3-D Plaxis
Continuation

Figures 5 and 6 show the deviatoric stresses distribution for 2.5 m column spacing with Hardening Soil Model. Figures 7 to 8 show principal stress direction prior and after installation of deep soil mixing. Dry Slope versus Wet Slope For the purpose of this paper, the slope is assumed dry. However, the parameters have been adjusted to simulate the condition where the ground water is approximately 2.0 m below road embankment. Comparison between the modified dry model and the original saturated model confirms a similar failure mechanism on the embankment slope for both models. On the saturated model, a deeper secondary failure mechanism is noted. The reported factor of safety for both models however are the same since the main failure is on the slope. It is noted that installing deep soil mixing columns have increased factor of safety by 20% to 40%. Modelling versus Reality The column spacing of 2.5 m was adopted in the final design. Whilst the factor of safety for the 2.5m spacing design appears the same as that of 4 m column spacings in the 2D models, the calculated shear stress significantly increases between the 0.6 and 2.5 m spacing and then appears to level out suggesting that the 2.5 m spacing design appears to be close to the threshold spacing for group effect. However, the same impact was not noted in the 3D models on the shear stress, but rather on the factor of safety. The deep soil mixing columns were constructed in November 2003. The remedial work has performed satisfactorily to date, and has undergone several events of heavy rainfall without developing any signs of further instability in support of the design adopted

Figure 5. 2D output - Deviatoric stresses for column spacing at 2.5m Hardening Soil Model

Figure 6. 3D output Deviatoric Stresses for columns spacing at 2.5m Hardening Soil Model

Conclusions
Figure 7. 3D Output - Principal stress direction prior to column installation Mohr Coulomb Model This study clearly demonstrates the 3D effects of soil column interaction. For closer column spacings of 0.6 m (two column diameters), interaction becomes more dominant, and 2D and 3D results coincide. For more typical column spacing of 2 to 2.5 m, there are significant differences between the 2D and 3D models which need to be understood. At wider spacings, the 3-D effects dominate, and the 2-D results may not be reliable. The key conclusion of this study is the need to understand the interaction of deep soil mixing columns with surrounding soil mass. Columns need to be designed so that they create an arching effect, and change stress distribution. Such change in stress distribution should be readily visible in model output, whether in 2D or 3D analyses, and should in most cases be reflected in computed factors of safety. Field observations suggest 3D effects of column interaction play a key role in the field performance of deep soil mixing columns in road remedial work. Figure 8. 3D Output - Principal stress direction for column spacing at 2.5m Mohr Coulomb Model

22

Recent activities
Erwin Beernink, Plaxis BV

Product updates
We are quite happy to announce that in the recent period, Plaxis and GeoDelft where able to achieve an update of our mutual product for Groundwaterflow analysis. With this update the stability and robustness of the program has been further improved. Most users of Plaxis V8 may not be aware, but the PlaxFlow calculation kernel has been serving smoothlessly their needs for stationary groundwaterflow analysis since the release of Plaxis V8 in 2002. In addition to that, the PlaxFlow software package for transient Groundwaterflow became available one year later. Especially after the number of PlaxFlow users kept growing and using the code for various different problems, our support desk experienced that some users had difficulty to finish some complex problems in the combination of flow and deformation analysis, and for some infiltration problems. Therefore last year it was decided to schedule an update of the code, and to build in the Unicode library for compatibility with Version 8. This upgrade of PlaxFlow, just improving the robustness, without adding much new functionality is one of the examples of continuous service and quality control for the Plaxis computer codes. We expect that especially those users that are working on problems related to time dependent interaction between slope-stability and groundwater flow will be pleased with the update of the program.
5 h(t) y 6

For detailed information on the V.I.Plaxis Service Program like pricing and subscription please contact us at sales@plaxis.com.

Plaxis Staff
We are pleased to announce that we extended our staff with Wendy Merks-Swolfs and Ronald Hordijk. Wendy studied Civil Engineering, and graduated at TU Delft, specializing in Structural Engineering and Computational Mechanics. Her main activity is the quality control of Plaxis products. Ronald studied computer science at the TU-Delft where he specialized in computer graphics. He will work on the output/presentation of all Plaxis products. Plaxis bv will continuosly strenghten its team and has currently additional positions available for; - Senior Software Development Engineer - Programmer Graphical User Interface - Numerical Geophysicist/Geohydrologis

Upcoming events
Last half year we had some novelties in our agenda, like user meetings in South East Asia and Russia (see photo), an advanced course in Mexico and the first course ever in South Africa. If you are looking for an opportunity to network with nearby Plaxis users or want to learn more about realistic simulations with Plaxis Products, please join us at the European Plaxis User Meeting 2006 or one of the other events. Final dates and locations will be posted on regular base at the agenda of our website. For more details on current planned activities, please check the backcover of this bulletin or the agenda on our website.

3 h(t) 0

h(t)

Furthermore we recently launched Plaxis 8.4 and 3DFoundation 1.6. Release notes on specific details can be found at the download area of our products. Plaxis Update Pack 8.4 includes bugfixes for all registered Plaxis V8 users and offers some new special extensions for V.I.Plaxis members.

V.I.Plaxis
The V.I.Plaxis Service Program is an additionalsubscriptionsystem on top of the traditionally Perpetual licenses. With the V.I.Plaxis Service Program you benefit from the latest releases of your Plaxis software, special extensions and support from Plaxis technical experts. The V.I.Plaxis Service Program protects your investment in geotechnical tools because it will keep your Plaxis softwareup to date. Each time a new version of your Plaxis software is released, it is automaticallyavailable. It cant get any easier than that. The special extensions are modular enhancements to Plaxis software products and are exclusively availableto V.I.Plaxis members. The extensionsprovide new functionalities, are fully compatible with the base product and are easy to learn.For members ofthe V.I.Plaxis Service Program the current available extensions for Plaxis 2D are: - A new Soil Model: Hardening Soil with Small Strain Stiffness - Soil Tests: Easy simulation of soil lab tests for Plaxis Soil models - Sensitivity Analysis; Analysis of parameters with the relative influences on the results - Parameter Variation; Upper and Lower Analysis for all possible combinations. June 2006, St. Petersburg, 1st Russian Plaxis User Meeting

23

Plaxis finite element code for soil and rock analyses

Activities 2006-2007
5 October 2006 Funderingsdag 2006 Ede, The Netherlands 28 October - 1 November 2006 Second National Conference on Geotechnical Engineering Wuhan, China 8 - 10 November 2006 European Plaxis User Meeting Karlsruhe, Germany 13 - 15 November 2006 Short Course on Computational Geotechnics Trondheim, Norway 22 - 24 November 2006 Pratique claire des lments finis en Gotechnique Paris, France 29 November - 1 December 2006 2nd International Workshop of Characterisation and Engineering Properties of Natural Soils Singapore 14 - 16 December 2006 Indian Geotechnical Conference 2006 Chennnai, India 11 - 13 December 2006 Plaxis standard course Jakarta, Indonesia 8 - 12 January 2007 Short Course on Computational Geotechnics Berkeley, California, U.S.A. 22 - 24 January 2007 International course on Computational Geotechnics Schiphol, The Netherlands 18 - 21 February 2007 GeoCongres 2007 Denver, Colorado, U.S.A. 21 - 23 February 2007 Course Computational Geotechnics Chennai, India 7 - 9 March 2007 UNSAT 2007 Conference Weimar, Germany 19 - 21 March 2007 Course Computational Geotechnics (German) Stuttgart, Germany 26 - 29 March 2007 International course for experienced Plaxis users Antwerpen, Belgium 25 - 27 april 2007 NUMOG X Rhodes, Greece 8 - 11 May 2007 16th Southeast Asian Geotechnical Conference Selangor Darul Ehsan, Malaysia 16 - 20 July 2007 PCSMGE 2007 Isla de Margarita, Venezuela 24 - 27 September 2007 XIV ECSMGE Madrid, Spain 10 - 14 December 2007 Asian Regional Conference Kolkata, India

Plaxis BV
PO Box 572 2600 AN Delft The Netherlands Tel: +31 (0)15 251 77 20 Fax: +31 (0)15 257 31 07 E-Mail: info@plaxis.nl Website: www.plaxis.nl

6005587