You are on page 1of 76
A 3D Numerical Simulation of Oblique Bending ina Steel Sheet Pile Wall A quantification of soil influence January 1997 EJ. Aukema A.G. Joling ate TU Delft sae recee ce eaiesine ao isrec bet Unvrstyotecmotagy nest saraon PRoFiL AED A.3D Numerical Simulation of Oblique Bending in a Steel Sheet Pile Wall A quantification of soil influence by EJ. Aukema AG. Joling Geotechnical Laboratory Report number: 416 January 1997 Supervisors: Foundation Engineering: Prof. it. A.F. van Tol ISPC profil ARBED: Dr-Ing. A. Schmitt Computational Mechanics: it. AE. Groen Foundation Engineering: ir. D.A. Kort Foundation Engineering: ir. LP, Oostveen Coordinator Hydraulic and Geotechnical Engineering: _ ir. K.G. Bezuyen A graduation project carried out at: In assignment of: Delft University of Technology ISPC profilARBED Recherches Faculty of Civil Engineering Division of Hydraulic and Geotechnical Engineering Foundation Engineering Section Division of Mechanics and Constructions ‘Computational Mechanics Section Preface This report is the result of our graduation project: A 3D numerical simulation of oblique bend- ing in a steel sheet pile wall. This project completes our study Civil Engineering at the Delft University of Technology and has been carried out under supervision of the Geotechnical Labo- ratory, Foundation Engineering Section. We are thankful for all the support from the section dur- ing this project. ‘The main goal of this study is to determine the effect of the transverse bearing capacity of the soil on oblique bending by a three-dimensional finite element analysis. Emphasis has been put on the generation of an appropriate mesh which is also useful for further investigations concern- ing oblique bending. ‘The report starts with a short introduction to the phenomenon oblique bending and some analyt- ical considerations. A description is given of the designed test case and of the used finite ele- ment model. Numerical results are presented and a contemplation is made how the designed ‘model is useful for further investigations. This project has been carried out in co-operation with ISPC profilARBED (Luxembourg) and the section Computational Mechanics of the Delft University of Technology, Faculty of Civil Engineering. We are grateful for the provided facilities and expertise which made this study pos- sible. ‘We would like to express our sincere gratitude to Dr-Ing. A. Schmitt, ir. D.A. Kort and espe- cially ir. A.E, Groen for their encouragements, stimulating discussions and guidance during the course of this study. Deli, January 1997 Evert Aukema Amoud Joling ‘Summary, conclusions and recommendations Summary In geotechnical practice frequently use is made of steel sheet ple walls for soil and/or water re- taining constructions. High demands are made with respect to the deformations, in particular in urban surroundings. Often these walls are composed of U- or Z-shaped profiles, whereby both profiles have their specific advantages and disadvantages. A common used variant is a sheet pile wall composed of 0 called double U-profiles. A double U-profile is made up of two single U-profiles which are connected by welding or crimping. A very specific property of the double U-profile is the rota- tion of the neutral axis. As a result of that, the sheet pile wall tends to deflect both in forward (lateral) direction as well as in sideway (transverse) direction. This phenomenon is called oblique bending, and it results in a loss of stiffness and strength (compared to a continuous wall) and thus larger deformations, Oblique bending goes together with slip in the free interlocks and a transverse deflection of, the sheet pile wall relative to the soil body. Oblique bending might be influenced by the trans- verse bearing capacity of the soil, ic. the soil reduces or even prevents the transverse deforma- tion of the sheet pile. Besides, the shear stress in the free interlock possibly reduces oblique bending. The main goal of this study is to determine the transverse bearing capacity of the soil. Therefore a three dimensional finite element model is made of a cantilever sheet pile wall with a dry excavation in front and without friction in the sliding interlocks. This is a "worst case’ mod- cl, ic. ifa sheet pile wall is modelled in this way, oblique bending can be observed most clearly. Emphasis has been put on the generation of an appropriate mesh which is also useful for further investigations regarding oblique bending. Use is made of the finite element system DIANA, ver- sion 6.1. First, the minimum and maximum stiffness of the sheet pile wall is determined by a numeri- cal simulation of two simply supported threaded double U-profiles with a distributed load. As plasticity in the sheet pile is not considered the stiffiesses are derived via elastic calculations. The maximum elastic stiffness is derived if oblique bending is prevented by fixing the free inter- locks (no in-plane deformation), the minimum elastic stiffness is derived if the in-plane defor- mation is free (slip in the free interlocks). The minimum stiffness of the double U-profile ap- peared to be 0.489 times the maximum stiffness. However, this ratio depends on the geometry of the used double U-profile and is called the reduction factor. The minimum and maximum elastic stiffness are used in two different 2D-calculations in which the minimum and maximum lateral deflections of the sheet pile are calculated. Besides in every 2D-calculation a distinction is made between a simulation with frictionless slip of the soil along the axis of sheet pile (5 = 0) and a simulation without slip of the soil along the axis of sheet pile (5 = @). If 5 = the passive earth pressure coefficient is several times larger than if 5 =0. Therefore the horizontal deflections of the simulation with frictionless slip are signifi- cantly larger. Furthermore a 3D-calculation is made in which the interaction between the sheet pile wall and the soil leads to reduced deformations of the sheet pile wall. The lateral deffections of the sheet pile wall in this calculation are compared with the results of 2D-calculations. The influ- ence of the transverse bearing capacity of the soil appears to be very small both calculations \with as well as without slip of the soil along the sheet pile wall With the help of analytical constitutive relations, the deflections and the stresses ca culated in the 3D-calculations, the moments in the sheet pile wall and the earth- and water pressure on the wall are derived. These results lead to the conclusion that the resulting water pressure on the sheet pile wall represents the main part of the total load at the back of the sheet pile wall. This is facilitated by the reduction of the earth pressure behind the wall to active pressure. Oblique bending, and thus the global behaviour of the sheet pile wall, is only influenced if the transverse bearing capacity of the soil is significant at the position where the horizontal deflections of the wall are relatively large. In case of a cantilever wall the largest horizontal deflections appear at the top of the wall. However, it was found that only at the lower side of the wall the soil stresses were high enough to generate a significant transverse reaction force. So, the influence of the transverse bearing capacity of the soil on the global behaviour of the wall appeared to be non significant. The use of anchors, struts or a decrease of the resulting water pressure can possibly increase the influence of the soil on oblique bending. Based on the derived moments, distributed loads, deflections and constitutive relations it is theoretically possible to determine the reduction factor of the sheet pile wall of the 3D- calculation analytically. The global behaviour of the sheet pile was accurately calculated, so a good indication of the value of the reduction factor (= 0.5) could be determined. However, it ap- peared that the results were not accurate enough in order to give an exact description of the soil stresses on a local level. Due to the limitations of the computer system the elements chosen were too large at certain points along and in the sheet pile wall. Therefore mesh variation is rec- ‘ommended in order to investigate the accuracy of the calculated results. In the Eurocode 3, Part 5 (lit{3]) the framework is given for the application of reduction fac- tors fp to the stiffness of single- and double U-profiles. First the reduction factor Bp is de- rived. If this reduction factor is applied to the stiffness of a continuous wall it gives the lowest stiffness. Six increments AB, are added to B pg, which represent various phenomena influenc- ing oblique bending. The numerical values of ABp, are expected to be given in the National Application Document, based on local design experience. Most of the effects in the six cate- gories can be implemented in the numerical 3D-model as designed in this study. Therefore the ‘numerical 3D-model is a useful instrument which can be helpful for the derivation of the numer- ical values of the reduction factors in the National Application Document. Conclusions ‘A three dimensional finite element calculation is a proper tool to investigate the phenomenon oblique bending in a steel sheet pile wall. In case of a high watertable the resulting water pressure represents the main part of the distribut- ed soil and water load at the back of the sheet pile wall and therefore the soil is not able to give a transverse restraint which significantly reduces oblique bending. ‘The numerical 3D-model is @ helpful instrument for the derivation of the numerical values of the reduction factors in Eurocode 3, Part 5. ‘The accuracy of the calculation results depends on the sizes of the elements used in the finite el- ement model. A good global behaviour of the finite element mesh does not imply a good de- scription on a local level. Recommendations ‘More 3D-calculations of fundamentally different cases will give a good insight in the way the soil develops its transverse bearing capacity. With the mesh that has been generated, the effect of other types of constructions and other soil conditions as well as measures to prevent oblique bending can be investigated. Possible other calculation cases could be: + A steel sheet pile wall with anchors or struts, including wailings, placed near the top or at several levels. + A steel sheet pile wall with a capping beam. + A steel sheet pile wall with the sliding interlock welded during the excavation, + A steel sheet pile wall without a resulting water pressure. + A steel sheet pile wall with friction in the sliding interlocks. In order to derive accurate and useful results of the calculation, an investigation of the influence of the mesh at the results is recommended. ‘Table of contents Preface Summary, conclusions and recommendations Table of contents Lntroduction 1.1 Aim of this study 1.2 General approach 1.3 The contents of the report 2 Analytical formulae for oblique bending 2.1 General analytical formulae for two directional bending 2.2 Analytical bending formulae applied at a double U-profile 2.3 The apparent stiffness of the 3D-sheet pile 3 Test case, dimensions and boundary conditions 3.1 Dimensions of the test case 3.2 Boundary conditions of the test case 4 Finite element modelling 5 Modelling the sheet pile 6 Modelling the soil 7 Numerical testing of the modelled sheet pile 8 Numerical simulations of oblique bending 8.1 Qualitative analysis based on the deflection curves 8.2 Quantitative analysis based on the deflections and the stresses as calculated in the 3D-simulations. 8.2.1 Theory behind the quantitative analysis 8.2.2 The distributed loads q(x) on the sheet pile wall 8.2.3 The moments M(x) in the sheet pile wall 8.3 Plots of stresses, strains and displacements of the 3D-calculation 9 Considerations on normative CUR and Eurocode publications 9.1 The Dutch CUR-publication 166 9.2 The Eurocode 3, Part 5, Draft 5 10 Conclusions and recommendations References Appendix 1: Calculation of soil loads from soil stresses Appendix 2: Calculation of moments of inertia of a quadratic profile Appendix 3: Example of DIANA Data- and Command-file 7 7 7 20 2B 26 28 3 33 35 35 37 39 43 50 50 52 36 37 60 63 1. Introduction For the formation of Eurocode 3, Part 5, an ECSC research project is being carried out, which is participated by ISPC ProfilARBED (Luxembourg), in order to provide knowledge on plastic design of steel sheet piling including plastic bearing capacity of steel sheet piles. One of the topics in this project is the phenomenon oblique bending. Oblique bending can appear in steel sheet pile walls constructed of so-called double U- profiles. These profiles are made up of two single U-profiles with the common interlock fixed by crimping or welding. As a consequence of the asymmetric cross-section of the double U-profile, the neutral axis is rotated, see Figure 1.1. A deformation of the sheet pile wall in lateral direc- tion, which will occur due to earth- or water pressure, car also result in a transverse deforma- tion. Neutral axis in oblique bending | Free interlock axial Zo | Fixéd imetock lateral Figure 1.1. Rotation of the neutral axis as observed in double U-profiles. In case of oblique bending, stip will occur in the free interlocks. As a result the strength and lat- eral stiffness of the sheet pile wall is reduced compared to a continuous sheet pile wall (ie. no slip in the interlocks), Oblique bending can be reduced by many factors, for instance friction in the free interlocks, embedment of the sheet pile head in a capping beam, anchoring etc. Moreover, the transverse restraint of the soil behind the sheet pile wall can reduce the transverse deformations. Within the framework of the ECSC research project three large scale bending tests on two double U-profiles have been carried out from February to April 1996 (by CRIF, Litge, lit{9)) ‘Two bending tests have been carried out on a naked profile and one test took the soil-sheet pile interaction into account. The profiles had a length of six meter and the load was introduced via ‘two stiffeners symmetrically placed at both sides of midspan (naked profiles) or via a load intro- duction plate on top of a sand bed, see Figure 1.2 and 1.3. In this way three different cases were tested: + A case in which oblique bending was influenced by the transverse bearing capacity of the sand bed, Figure 1.32 + A case in which oblique bending was prevented by a transverse restraint, Figure 1.3b. + A.case in which free oblique bending occurred, Figure 1.3c. Load introduction plate @ © Figure 1.2. Cross-section of the profiles tested by CRIF, Lidge. () (©) Figure 1.3. Top views and a side view of bending tests carried out by CRIF, Liége ‘The interpretation of the bending tests is extensively described by Kort and van Tol in the report Analytical and 2-D numerical considerations on large scale oblique bending tests (lt{8)). In this chapter only the findings are given which are leading to this study. From the results of the two bending tests on a naked profile a relation could be made between the transverse restraint force and the horizontal deflection at midspan, see Figure 1.4a. This relation is linear if the vertical load F,, at the sheet pile is constant. In case of transverse restraint bending the section properties / and W of a continuous wall apply whereas in case of ‘unrestraint bending reduction factors R, and Rw should be applied. These factors depend on the ‘geometry of the steel sheet profile and are in the range of Ry = 0.6 and Ry = 0.7. By means of 2D-finite element simulations the relation is derived between the transverse bearing capacity of the soil of the bending test with the sand bed and the horizontal deflection at midspan, see Fig- ure 1.4b. The two figures were plotted in the same diagram which gave an equilibrium point, see Figure 1.4c. If the sand is capable of bearing a part of the restraint force, for example } F, the reduction factor Ry increases from Ry =0.7 to Ry =0.8. From the equilibrium point between the beam line and the soil resistance it can be derived that R, increases from Ry = 0.6 to R =0.75. It is emphasized that these relations are derived from a theoretical analytical study. Based on the 2D-calculations it was concluded that the transverse bearing capacity and stiff- ness of the soil can have a significant influence on the stiffness of the sheet pile. Based on the test results though the transverse bearing capacity appeared to be negligible small. Possibly these differences occurred due to the fact that the shear deformation of the soil adjacent to the sheet pile in oblique bending is a typical 3D-problem. Fy Fi, Ry=1 equilibrium Ry =08 Ry =01 Ry=l Ry = 0.75 @ (b) © Figure 1.4. Relations between the transverse restraint force and deflection for the sheet pile and soil derived by Kort and van Tol As a logical follow up, ISPC ProfilARBED sponsored our study in order to design a numerical 3D-finite element model in which the influence of the transverse bearing capacity of the soil on oblique bending of a steel sheet pile wall can be derived. Due to the fact that there was no experience available concerning calculations with an appro- priate 3D-finite element model in which the soil is modelled elasto-plastic in combination with a correct steel sheet pile mesh, an enormous effort of this study was put in making the available 3D-finite element system DIANA, version 6.1 accessible for this problem. 1.1 Aim of this study Aim of this study was to find out whether the soil is able to prevent a sheet pile wall from oblique bending. In order to investigate the transverse load bearing capacity of the soil behind a cantilever wall, a three-dimensional finite element analysis has been carried out. Emphasis has been put on the generation of an appropriate mesh which is also useful for further investigations concerning oblique bending. The finite element system DIANA has been used for the numerical analyses. 41.2 General approach First a test case of an excavation in front of a sheet pile wall has been designed. With the help of ‘2D-calculations of a vertical cross-section perpendicular to the sheet pile wall it was determined whether the mesh boundaries, dimensions, boundary conditions, element types, constitutive models and material parameters were chosen correctly. Also the 2D-mesh was used to determine how the "Phased Analysis” tool of DIANA is applicable for simulation of an excavation in a geotechnical calculation Based on the test case, a 3D-model of an infinite repeating sheet pile wall has been gener- ated. Since no appropriate 3D-mesh generator was available, several programs have been written in order to generate meshes efficiently. The capacity of the computer system restricted the size of the 3D-mesh (amount of elements). Obtaining an appropriate 3D-mesh was a matter of trial and error. In order to determine the influence of the soil on oblique bending and to determine the trans- verse bearing capacity of the soil, three different cases have been calculated: + A situation in which it is assumed that no oblique bending in the sheet pile wall occurs and hence the sheet pile wall has its upper limit lateral stiffness. A continuous sheet pile wall with all interlocks fixed is modelled. + A situation in which it is assumed that oblique bending is unrestrained. The soil has no transverse bearing capacity. Hence the sheet pile wall has its lower limit lateral stiffness, + A situation in which it is assumed that oblique bending is influenced by the transverse bear- ing capacity of the soil. In this case the sheet pile wall has a lateral stiffness between the upper and lower limit. Note that the lateral stffiness of the sheet pile wall changes because of the boundary conditions (interlock- and soil behaviour), and not by changing the geometry or material parameters of the individual sheet piles. The 3D-model has been used for two different simulations of an excavation in which oblique bending is influenced by the transverse bearing capacity of the soil. A distinction has been made between frictionless slip and no slip of the soil along the sheet pile wall in axial direction. So, the sheet pile is either perfectly smooth (5 = 0) or perfectly rough (5 = 9) in axial direction. In order to determine the behaviour of a sheet pile wall without the soil influencing oblique bending (lower limit stiffness), and a continuous sheet pile wall without oblique bending (upper limit stiffness), use was made of the 3D sheet pile without the soil. With the help of 3D numeri- cal bending tests, the “apparent stiffness” of the sheet pile in free oblique bending and without oblique bending (transversely restrained) was determined. The apparent stiffness is the stiffness calculated on the basis of lateral (i.e. out-of-plane) displacements as if the sheet pile were a line- clement. Naturally in case of the transversely restrained sheet pile there are only out-of-plane displacements. ‘The apparent stiffness has been used in 2D-analyses of a vertical cross-section perpendicular to the sheet pile wall with an excavation in front of a sheet pile wall. These 2D-calculations gave the lateral displacements of the sheet pile wall with the upper limit and lower limit stiffness. These displacements were compared to the 3D-results of the excavation with the soil influencing oblique bending. In this way the influence and the transverse bearing capacity of the soil was determined. In Figure 1.5 the general approach is recapitulated in a flow-scheme. Lower limit stiffness Upper limit stiffness of S.PLW. (apparent stiffness) of SPW Ne ‘Num. bending tests t 3D - model of D.UP. ! ! 3D - model of infinite SPW. in soil (based || ——>] con the testcase) | sew SPW behavourif Gomes cee the soi reduces soa ce’ oblique bending sas coninuous SPW. | ‘Transverse bearing 2D - model of infinite S.PW. 2D - model of infinite S.PW. capacity of the soil ! | Maximum reduction 2 No reduction 1 Figure 1.5. Flow scheme of the General Approach ( $.P\W. = Sheet Pile Wall, D.U.P. = Double U-profile, r = reduction factor), 1.3 The contents of the report In Chapter 2 the general analytical formulae for two directional bending are applied to a double U-profile. Based on these formulae the definition of the apparent stiffness is derived. In Chapter 3 the designed test case, which is implemented in DIANA, is explained including the dimen- sions and the boundary conditions. Chapter 4 deals with the implementation of the test case in DIANA. The type of elements are discussed as well as the connection between the soil- and sheet pile elements, mesh refinements and the application of the "Phased Analysis’ tool of DIANA in case of a staged excavation. In Chapter 5 the material properties and the shape of the 10 modelled sheet pile are given, in Chapter 6 the constitutive model and material parameters of the soil (sand) are given. In Chapter 7 the results of the numerical bending tests are given, which are used in order to derive the stiffness and behaviour of the modelled stee! sheet pile. In Chapter 8 the results of the 3D-simulations of the test case are discussed, including the derived reduction factor 'r’. In Chapter 9 the Dutch CUR-publication 166 and the Eurocode 3, Part 5 will be dis- cussed (concerning the application of reduction factors on the moment of inertia) based on the results of this study. Finally, in Chapter 10 the conclusions and recommendations are given. i 2. Analytical formulae for oblique bending, In order to get a comparison between the 2D- and 3D-calculations, the stiffness parameters of the "numerical" 3D-sheet pile had to be transformed in the 2D-calculations. As these parameters differ from the parameters which can be calculated by the geometrical dimensions, the stiffness parameters of the “numerical” 3D-sheet pile are obtained by using general analytical formulae for two directional bending. The analytical formulae are also used for the evaluation of the numerical results of the 2D- and 3D-calculations. The resulting loads and moments on the sheet pile wall can be derived from the displacements by using constitutive relations. 2.1 General analytical formulae for two directional bending. In Figure 2.1 and 2.1b the definition of all variables and notations are given (lit{11]). Figure 2.1 a: Coordinate system as used in analytical formulae. bs Coordinate system and prin- cipal directions as used in analytical formulae (with working plane m and k of M resp. x). The constitutive relations for two directional bending are: (i elle] ES W , 12 A Pw, Fw, and xy =~ ve oS Oe on Inversion leads to the following relation: ky 1 Te ~Tyz |{ My Ke] EUyyl—Toly) |—ly by || Me From the Mohr circle (Figure 2.2) can be derived: L ye tan(2@) = ;~——"*— 3 Clyy — Ta) in which a gives the rotation of the principal directions compared to the chosen coordinate system. j | Ire Figure 2.2 Mohr circle in which A gives the rotation of the neutral axis compared to the chosen coordinate system. ‘Assum in which 7 gives the rotation of the working plane m compared to the chosen coordinate system. Note that cr only depends on geometrical properties, # depends on both geometrical proper- ties and the load case, and 7 depends on the load case. 13 2.2 Analytical bending formulae applied at a double U-profile In Figure 2.3a and 2.3b the definition of all variables and notations are given. Figure 2.3 a: Coordinate system as used in analytical formulae. 6: Coordinate system and prin- cipal directions as used in analytical formulae If no oblique bending occurs, applies: Note that = a only applies if y = a, ie. the vectorial sum M of M, and M, works in E-direction. ‘The restraining moment M, can be provided by either the transverse bearing capacity of the soil or by a distributed shear force in the interlocks. ‘An analytical relation can be made between the transverse distributed load q,(x) of the soil at the sheet pile, and the moment M(x) in the sheet pile caused by q,(x): omy ee as) Also an analytical relation can be made between the distributed shear force T(x) in the "free" interlocks, and the moment M(x) in the sheet pile caused by T(z): aM, ax =2bT (x) in which b is the distance between the interlocks, so 2b is the distance between the "free" inter- locks. It should be emphasized that the two previous relations are based on Bernoulli's beam theory. Especially the relation between M,(x) and T(x) should be used with care. It can only be applied to derive an order of magnitude of T(x) in a continuous wall. 2.3 The apparent stiffness of the 3D-sheet pile In order to get a comparison between the 2D- and 3D-calculations, the stiffness parameters of the “numerical” 3D-sheet pile had to be used in the 2D-calculations. These parameters are obtained by a numerical simulation of bending tests at the 3D-sheet pile. With the help of these bending tests the “apparent” stiffness of the 3D-sheet pile is deter- mined. This is the stiffness that is observed when only the out-of-plane bending of the sheet pile is measured, as we do in 2D calculation In this paragraph the deflection formulae and apparent stiffness for a simply supported beam with uniformly distributed load are derived. eM, aM, an 4 Oye Assume: — a With the constitutive relations for two directional bending it can be found that for a simply sup- ported beam with uniformly distributed load the formulae for the deflection at mid span (x=h Dare: 5 5 Ie pay 2 Sgt te 384 BOgta—taly) 384 Eyl Tul) mo 5 Le 5 4, 35 pe 384%" EOF, Tal) * 384 BUT, w= Note that the term 5/384qi* may be replaced with any other term belonging to the loading con- dition. ‘The deflection formulae for two test cases are derived: + Transverse restrained bending on a double U-profile. + Transverse unrestrained bending on a double U-profile, 15 In case of transverse restrained bending the boundary conditions are: wy20 > Ky This gives the following deflections at mid span (x=): It should be noticed that this relation is general. In order to obtain a full restraint, the restraining force should be congruent with the load, but a factor I,./1,, larger. In case of transverse unrestrained bending the boundary conditions are: qy=0 => My This gives the following deflection formulae at mid span (x = 3 D) 5 s we 39g at ECoyla—Tpl, sels) Sahn 384 Byla —Tylo) ‘The definition of the apparent stiffness is: This gives the well known (2D) relation for the deflection formula: Sat 384 El, ‘The derived apparent stiffness EJ‘, will be used in the 2D-calculation in which it is assumed that no oblique bending in the sheet pile wall is activated. E7’, is the observed bending stiffness of the sheet pile wall, which is the lower limit stiffness in case of unrestrained bending. 16 3. Test case, dimensions and boundary conditions. In order to study the phenomenon oblique bending, a test case is designed. This test case is a model of an excavation of a soil body (sand) in front of a cantilever steel sheet pile wall and is used for implementation in a finite element model. As the model has finite dimensions, it is essential to applicate correct boundary conditions in order to obtain a proper model. 3.1 Dimensions of the test case. ‘The dimensions of the modelled soil body and the position of the cantilever wall in the soil body are of such magnitude that if the boundaries are chosen further away from the sheet pile wall, the behaviour of the wall does not change, see Figure 3.1 12m It lateral transverse — Exe. depth = 0.9 m Exe. depth = 1.8m Exc. depth= 2.7 m Water pressure fon sheet pile wall 15m Sand Figure 3.1. Dimensions of the test case. ‘Three excavation phases are simulated. Each phase represents 0.9 meter dry excavation in front of the sheet pile wall. A drained condition is assumed which means that no excess pore pres- sures are generated. The water content of the sand does not have any effect on the deformations except via a load at the sheet pile wall. The resulting water pressure at the sheet pile wall is rep- resented by a distributed load which changes with the depth of excavation, see Figure 3.1. The dimensions of the model in transverse direction will be explained in the next paragraph. The geometry and material parameters of the sheet pile wall and the material parameters of the soil will be given in Chapter 5 and Chapter 6, respectively. 3.2 Boundary conditions of the test case. ‘The main assumption is that the sheet pile wall has an infinite length in transverse direction. In this way the effects at the end of the sheet pile wall are neglected, which allows to assume a periodic behaviour. One period takes up the width of one double U-profile, so every volume with a width of a double U-profile meets the requirements of periodic behaviour. A volume between the fixed interlocks is chosen in view of the implementation in a finite element model ‘A top view of the volume with the applied periodic boundary conditions is shown in Figure 3.2. a7 Fixed interlock Free interlock Fixed interlock Figure 3.2. The displacements u,v ,w and rotation @ on both sides of the considered volume are in order to enforce periodic behaviour. With these boundary conditions a semi plane strain model is created. The fact that the displace- ments u,v, W and rotation @ are equal at both sides of the considered volume, does not imply that these quantities are equal to zero The boundary conditions at the other planes of the model are depicted in Figure 3.3 Figure 3.3. Boundary conditions at the front (= left), back (= right) and bottom plane of the model. [At the front the model is supported in transverse and lateral direction. In this way the model behaves as if the front is the symmetry axis of a building pit. At the back of the model the same boundary conditions are applied. The bottom of the model is supported in all directions. Figure 3.4 gives a three dimensional view of the test case. 18 transverse 6.0m lateral axial Figure 3.4. Three dimensional view of the test case. 19 4, Finite Element Modelling This chapter deals with the consequences of using of the finite element method for calculation of the test case as described in chapter 3. Also, the limitations imposed by the computer system ‘on the modelling process will be discussed. For finite element modelling in DIANA, choices had to be made concerning the element types. Because bending of the sheet pile can not be properly calculated with four noded shell elements (lit{12}), eight noded shell elements have been used to model the sheet pile. For con- nectivity reasons, this leads to the use of twenty noded brick elements for the soil. Both ele- ‘ments are shown if Figure 4.1. Figure 4.1. Bight noded shell element and twenty noded brick element. For modelling friction and slip in the contact surface between soil and sheet pile, interface ele- ments had to be used, in this case eight+eight noded interface elements. Because it was thought that use of these elements was riskful regarding the numerical stability, and very few experience was available, it was chosen not to use interface elements, but to use ‘tyings’ in the contact sur- face, as will be explained later on. In this way, two limit cases can be treated, namely a perfectly smooth sheet pile (5 = 0, frictionless slip) and a perfectly rough sheet pile (6 = ¢, no slip - soil attached to sheet pile). In the contact surface between soil and sheet pile, a one-on-one relation between brick ele~ ments and shell elements is needed. This implicates a one-on-one relation between soil nodes and sheet pile nodes as well. This is needed to assure that all non-boundary nodes are connected to at least one element. Experiments have been carried out where this one-on-one relation did not exist. This revealed improper displacement behaviour of nodes that were connected to only one element. ‘As the shape functions of the Mindlin shell elements and the brick elements are compatible, the translational degrees of freedom are compatible. Obviously, the rotational degrees of free- dom of the shell elements can not be tied to the brick elements, as the brick elements do not fea- ture rotations. Given a one-on-one element relation, it follows that the amount of brick elements bordering the sheet pile is governed by the amount of shell elements in the sheet pile, needed for a proper calculation of sheet pile bending On the other hand, computer capacity was a limiting factor. Large amounts of small ele- ‘ments lead to long calculation times and large disk space usage, as the number of data increases rapidly with the number of elements. All data, appropriate for analyses of a particular FE~ model in DIANA, is saved in a central database file called "FILOS file”. The operating system 20 IRIX version 5.3) was unable to handle a FILOS file larger than 2 Gigabyte. Furthermore, the size of the FILOS file increases during the analysis. It appeared difficult to predict the size of the FILOS file in advance, based on the amount of nodes and elements. Therefore, obtaining a usefull 3D-mesh was a time consuming cycle of trial and error. An extra consideration was the total calculation time. A mesh that can be handled by the computer system, but with a calcula- tion time of a few weeks for the complete analysis is not very useful either. ‘As most action in the soil is concentrated near the sheet pile and because the number of ele- ments had to be kept as small as possible, it was desirable to have smaller elements near the sheet pile and larger elements further away. For this purpose, mesh refinement zones had to be created where a transition from small too large elements takes place. For mesh refinement, two possibilities were considered, as shown if Figure 4.2. Figure 4.2. Mesh refinement options In the first case (left) transition takes place at a line, where half the mesh lines stop. This implies that also half the nodes are connected to just one element, allowing this node improper displace- ment. For four noded elements with a linear displacement over the element, this problem can be solved by imposing a linear dependency between the displacements of this “free” node with its neighbours. For eight noded (quadratic) elements this solution is not available, and improper displacements can not be prevented. Therefore, this way of mesh refinement was excluded from further consideration. In the second case (right), there are no abrupt endings of mesh Lines and no "free" nodes. The only limitation for this mesh refinement is that the angles in the transition elements should not be too small. ‘The DIANA mesh generator supports only regular 2D meshes. Refinements and 3D meshes are not supported. To simplify mesh building, several additional mesh processing programs were made, for purposes as: + 2D mesh refinement + 3D mesh out of 2D base + 3D mesh refinement + 2D mesh slices out of 3D mesh + joining meshes For imposing the boundary conditions, as described in chapter 3, the DIANA feature ‘tyi has been used. With this feature, degrees of freedom of one or more ‘slave nodes’ can be tied’ to that of another node, the "master node’. Any linear dependency can be applied for this tying, but in this study only equal displacement relations have been used. Tyings have been used to 21 model periodic boundary conditions for simulating an infinite wall, and at the boundary surface of soil and sheet pile. For modelling periodicity, the two ‘side’ planes perpendicular to the length of the sheet pile wall (these are the planes used in the 2D calculations) have to be tied for all degrees of freedom. For twenty noded brick elements these degrees of freedom are transla- tions in three directions. For the contact surface between soil and sheet pile, two conditions have been simulated; a perfectly smooth sheet pile (frictionless slip) and a perfectly rough sheet pile (no slip - soil attached to sheet pile). ‘Tyings for perfectly rough conditions imply equal dis- placements in all three directions for adjacent soil and sheet pile nodes. This does not give any problems. Tyings for perfectly smooth conditions imply equal displacements in directions per- pendicular to the sheet pile surface and non-equal translations (sliding) along the sheet pile. This sliding along the sheet pile can be divided into axial sliding and non-axial sliding. The non-axial sliding leads to difficulties in comer nodes of the profile. As tyings are always applied perpen- dicular to the sheet pile surface (to avoid soil nodes from moving "through the sheet pile), there are two tying directions in corner nodes, as there are two normal vectors on the sheet pile sur- face. This means that no non-axial slip remains possible in comer nodes of the profile, as shown in Figure 4.3. No slip possible f 4A i» aa / joa transverse T : lateral Figure 4.3. Blocking of slip in comers of the profile would lead to peak stresses. If non-axial slip was allowed in other nodes along the sheet pile surface, this would lead to peak stresses in the comer nodes, because they would have to take care of all slip stresses of the other nodes. Therefore it has been assumed that only axial slip can occur, and non-axial slip is pre- vented by tyings in both lateral and transverse direction. The physical reality of this assumption seams to be reasonable, as the soil is "locked up’ by the sheet pile. The assumption possibly leads to a small overestimation of the transverse force from the soil on the sheet pile. For simulating the different calculation phases, use has been made of the ‘Phased Analysis’ tool of DIANA (lit(4]). Hereby it is possible to simulate the excavation in front of the sheet pile wall. First the soil stresses due to the gravity load are generated. Thereafter the excavation is simulated with steps of 0.9 meter. This is the height of one element, but more elements is also possible. ‘At the start of the first excavation step the first elements are removed. The load of the removed elements at the surrounding elements is replaced by an external force. During the cal- culation phase, which can exist of up to several hundred iteration cycles, the external force is removed step wise. All the stresses due to forces as calculated in previous phases will disappear stepwise, but the stresses due to forces which are applied in the present phase will build up step- wise. If nothing changes in the situation, ic. in the present phase the same forces are applied as in the previous phase, the decreasing and increasing forces remain in balance. 22 5. Modelling the sheet pile ‘The sheet pile used in this study is a simplified version of the PU8 profile from Profil Arbed it{5), shown in Figure 5.1. In this figure the rotation angle of the neutral axis a is shown, as well as some other measures of the profile. Figure 5.1. The PUS profile with rotation of neutral axis o and mass center of single profile at .<). The simplification is needed because the soil elements must follow the shape of the sheet pile elements, and every sheet pile element must have a matching element in the adjacent soil, as described in chapter 4. For this reason, a well fitting model shape for the sheet pile model as shown in Figure 5.2a is not useful as this would result in too many small adjacent soil elements. Therefore, the simplified model shape as shown in Figure 5.2b has been used. This model has angles of 45° instead of 48.6° but the flange length is the same. 48.6 457 Zz @ ®) Figure 5.2. Two possible shapes for modelling of the PUS profile. ‘The aim was to determine the thickness of the shell elements over the profile in such a way that, in spite of the different shape, the model has elastic properties similar to that of the PUS profile. ‘The main difficulty resulting from the simplified shape of the model is to simulate the amount of ‘mass concentrated in the interlock. As the sheet pile is modelled with eight noded shell elements based on quadratic interpolation, a nonuniform thickness varies quadratically over the element area as well. thas been assumed that the flange thickness is uniform and that the starting thickness of the body of the profile (in the comer of the profile) is equal to the flange thickness, so there is no change of thickness in the comer of the profile. The area of the body is built up of a box and of two parabolic parts, as shown in Figure 5.3. As the thickness is not supposed to decrease towards the interlock, and to place as much mass as possible near the interlock, the parabolic part of the thickness is assumed to have a tangent parallel to the axis of the profile in the comer of the profile. 23 Figure 5.3. Thickness variation over the profile ‘The moments of inertia of the model profile can be calculated from the known moments of iner- tia of its composing parts, using transformation rules for translation and rotation. There are two parameters that can be varied, namely the thickness of the flange and the thickness in the interlock. The profile that turned out to have the best parameter fit is shown in Figure 5.4 Mem __32em 28cm ___32em_ Idem Figure 5.4. Dimensions of the cross-section of the used double-U profile. However, it tumed out that it was not possible to keep the z-coordinate of the mass-center of a singe U-profile (distance ¢ in Figure 5.1) of the profile as small as necessary. As Tyzdouie = 2- Acinge- B-C this means that /,_ is overestimated, while all other properties of the PUS-profile could be matched quite accurately. This results in a higher rotation angle of the neu- tral axis c, a lower value for the smallest eigenvalue J, and a higher value for the largest eigen- value J. In Figure 5.5, Mohs circle is drawn for both the PU8 profile and for the approxi- ‘mated profile used in this study. Because J,, and J, are the same for both profiles and only the Iyz is different, this yields two concentric circles. 24 Table 5.1. Properties of (double) PUS Profile and approximation, PUS | Approx. Asingie (cm?) | 69.5 69.3 (cm) 8.15 8.66 Ly (om) 183100 | 183143 1. (em) -33986 | -35990 1, (cm) 13953 | 13953 Ti (cm* 7367 6615 T (cm) 189673 | 190481 @(’) 10.95 11.52 | 1 Figure 5.5. Mohr circles: inner circle for PUS and outer circle for approximation. The sheet pile is assumed to be fully elastic during all calculations in this study. Plasticity in the sheet pile is not the subject of this study, and would also require more detailed meshing of the sheet pile, which is hard to combine with the FE soil model, as described earlier. The elastic parameters of the sheet pile used in this study are Young's modulus: E=2.1-10!'N/m? Poisson's ratio v=0.3 6. Modelling the soil ‘A 3D-view of the finite element mesh of the soil body after 2.7 m excavation is shown in Figure 6.1a. Twenty-noded brick elements with full (27-point) quadrature are applied to model the soil ‘mass. For computational convenience it would be advantageous to apply a reduced (i.e. 8-point) quadrature rule. However, calculations revealed unstable element behaviour with reduced inte- gration and was therefore excluded from further consideration, Drucker-Prager =~ Zionkiowicz & Pande fitto Lade & Duncan van Eekelen ~~~ Moby-Coulomb 50 deg. \\, A SN fa) ) Figure 6.1. (a): 3D-view of the utilized finite element mesh. (b): Several possible fits to the Mohr-Coulomb yield criterion. The soil body is uniform, and is assumed to obey a linear elastic-ideal plastic frictional material Jaw in which the yield surface is defined by a smooth fit to the Mohr-Coulomb criterion (van Eekelen 1980, Figure 6.1b.) and in which the flow direction is determined from a Drucker- Prager plastic potential (lit{6}&(7). In a finite element framework the stress paths are not known in advance and they are cer- tainly not constant. Implementation of the Mohr-Coulomb criterion in a fully three-dimensional continuum poses some difficulties which adhere to the multi-surface character of the Mohr- Coulomb yield criterion. Therefore, the objective is to define a model which is continuous, and is able to accurately represent the strength of sands under various triaxial stress paths. The model of van Eekelen (1980) has been used, because it provides a close fit to the Mohr- Coulomb criterion, Several 2D-simulations were made in order to compare the results of both material models. ‘These results showed no significant differences between the Mohr-Coulomb and van Eekelen ‘model. However, the simulation with the van Eekelen model is much more robust. ‘The material properties which have been assumed for the sand are: Young's modulus: | E=20MPa Poisson's ratio 33 effective weight: y’ =10KN/ mn? 26 friction angle: o=35° dilatancy angle: y= 5” cohesion: c=1 kPa As a drained analysis is assumed the water content does not have any effect on the deformations except via the water pressure at the sheet pile wall. In order to generate realistic effective stresses in the soil, the “under water’ density of the soil is used and the resulting water pressure ‘on the sheet pile wall is modelled as an external distributed load. (See Chapter 3) 27 7. Numerical testing of the modelled sheet pile ‘The 3D FE-model of the sheet pile has been tested numerically, to derive the upper and lower limit stiffness for use in 2D calculations, and to compare the numerical results with the analyti- cal bending solution. At the time the sheet pile model was finished, the complete (soil and sheet pile) model was not ready, but the modelled sheet pile was already tested numerically. As a result of this, most of the numerical sheet pile test have been carried out on a different sheet pile than the one used in the final model. The number of elements over the length of the sheet pile varied with every new soil model, until the model could be handled by the computer system available. After finishing the soil model, the final sheet pile used in this model had to be tested less intensively, because earlier tests had already given insight in the numerical behaviour of the modelled sheet pile. The sheet pile used in the earlier numerical tests has a length of 6 m and mesh elements with ‘an axial length of 0.2 m, while the sheet pile used in the soil model has a length of 9 m and ‘mesh elements with an axial length of 0.9 m. To judge the numerical behaviour of the sheet pile, 4-point bending tests have been carried out, similar to the ones done in Liége within the framework of the ECSC research project. Two tests were simulated, an unrestrained test and a restrained test (oblique bending impeded). The model consists of two double U-profiles with stiffeners at both ends (0 m and 6 m) and at load introduction (2.6 m and 3.4m). The sheet pile are supported at both ends. The load is applied by a line force load over the stiffeners. In the restrained test, the restraints are located at the ends and at load introduction, as shown in Figure 7.1b. [—— | C~ i (a) b) Figure 7.1. Load and supports of 4-point bending tests. (a) unrestrained, (b) restrained. ‘The analytical formulae for the deflections at load introduction for this case are: 1, 1 _ re, Lag _ 4 lan Tyly) 8 °° BU Ly 1 Fa Ly "Byte Taly) 8° Edyla—Tyly) Where ‘a’ is the distance between the support and the load introduction, as shown in Figure 7.1 and 'F,” is the value of one Force Load (total load is 2F), and the restraining force F, in each support is: 28 ‘The numerical calculations show the following deformations: Figure 7.2. 3D, Top-, side- and frontview of deformed mesh of unrestrained 4-point bending test (0.2m mesh - displacements scaled up) To get an idea of mesh dependency of the results, both test have been repeated with an element length of 0.2 m. as well as with an element length of 0.5 m. This difference in element length is only in axial-direction and not in the other directions. The results are summarized in Table 7.1 Table 7.1. Deflections and reaction forces (at load introduction) of 4-point bending tests, F, = 120kN [4-point bending test | Analytical | Numerical | Numerical F,=120kN Solution | 0.2m Mesh | 0.5m Mesh Unrestrained w, | 1432em | 1428em | L610cm w, | 7.2850m | 7.322cm |_7.540cm Restrained Fy | 309.5KN | 2824KN | 281.7KN w, | 3592¢m | 3.963¢m | 3.76cm In the unrestrained case, itis observed that both w, and w, deflections are higher in the numeri- cal simulations than in the analytical formulae. The finite element calculation shows @ weaker behaviour than the analytical solution. The restrained test shows a weaker behaviour as well, and the difference with the analytical solution is somewhat larger. These differences can be explained by the fact that the sheet pile is not a line element and that it does not have to obey the Bernoulli hypothesis (both are assumed in elastic bending the- ory). The main difference with a line element is that the sheet pile is able to deform internally, i.e. the profile of an intersection plane can deform as well. It is reasonable to assume that these internal deformations increase if the sheet pile is restrained from oblique bending. ‘The idea of these internal deformations was affirmed by other tests, in which there were only stiffeners at the supports, and not in the middle. These tests were as well carried out on a sheet pile a length of 6 m and mesh elements with an axial length of 0.2 m. In these tests, the in-plane deformations of the cross sections are clearly visible in a view in 29 axial direction, as shown in Figure 7.3. The size of these internal deformations appeared to ‘depend upon the type of loading. The in-plane deformation is the largest for a line force load, smaller for a distributed load and almost invisible for a moment loading. This leads to the assumption that the in-plane deformations are caused by shear forces. Figure 7.3. Internal deformations for three different load cases with deformed mesh views in axial direction. To test the effect of these in-plane deformations, tests have been carried out in which in-plane deformations have been suppressed by demanding equal in-plane displacements Ax and Ay for all nodes in the same z-plane (same axial coordinate). When in-plane deformations are sup- pressed the sheet pile behaves stiffer. With this test setup, the effect of different interlock conditions have been tested as well. This is also useful to test the difference between one double U-profile and an infinite series of double U-profiles. ‘The following three interlock conditions were tested: + completely free interlock, no tyings or supports (‘free’). + interlock with axial sliding and rotation possible, but translations perpendicular to the pile axis prohibited (‘slide’). + interlock with all translations and rotations fixed to the other interlock by means of tyings fixed’), The first condition is only possible for the outer interlocks of a double U-profile (the middle interlocks are always connected), the other two conditions can be imposed on both the middle interlock the outer interlock of a double U-profile. A certain case is denoted by two keywords for the middle interlock and for the outer interlock (c.g. 'fixed-slide’). As expected, there was no difference between the ‘fixed-slide’ and the "slide-fixed’ case, except for a reversal of the oblique bending direction. Less expected however was the fact that there was no substantial difference between the 'fixed-slide’ and the ‘fixed-free’ condition. This means that one double U-profile (where the outer interlocks are free to move apart or towards each other) behaves nearly the same as a double U-profile in an infinite wall (where all sliding interlocks are forced to bend the same way). This also means that no force perpendicular to the axis is activated in the interlock by Bernoulli bending. From the deflections at midspan, the apparent moments of inertia of the numerical sheet pile have been calculated in accordance with the elastic formulae: 30 Force load: Distributed load: Moment load: m= & Eitan Table 7.2. Apparent moment of inertia of the profile for different load cases “Anajizal Reus | Nemercal Reus | Numerical Revue (no n= plane efomatons of intersection) | Wee | Leet | Wenn | Lene | Wem Foca Load type (em) | (em? | cm) _| (cm (cm) (cm® F=180kN/m 6.728 | 6880 | 7.228 | 6404 6933 | 6676 | 50 kN/m? 7.008 | 6880 | 7.363 | 6548 | 7.201 | 6695 | M=208kNmim__| 7.774 | 6880 | 7.958 | 6721 | 7.920 | 6753 _| The question is which apparent moment of inertia is most representative for a sheet in pile soil conditions. Because the soil inflicts a sort of distributed load, the apparent moment of inertia following from a distributed load was considered to be most representative. Finally, the sheet pile used in the soil model (with an axial element length of 0.9m), had to be tested with a distributed load. For this purpose the 9m sheet pile was equipped with stiffeners at both ends, similar to the other tests. This yielded an apparent moment of inertia J” = 6704 cm‘. ‘This value has been used as the lower limit in the 2D calculations. For the calculation of the 2D upper limit moment of inertia, a simulation has been carried out in which both interlocks were fixed. In this case the outer interlocks were fixed to each other for all degrees of freedom by ‘means of ‘tyings’. This yielded an apparent moment of inertia of 1° = 13718 cm‘. The relation between the apparent moment of inertia J” and a reference moment of inertia J, is most com- monly expressed as a reduction factor: 1 he The reference value Iyer is the upper limit moment of inertia following from the same type of calculation, i. r Tanase = and ure T, zum aperint Values for J", Ig and r are summarized the following table. The values from the bottom row Numerical’) have been used as upper and lower limit moments of inertia in the 2D calcula- tions. 31

You might also like