Appendix G 3D FLAC Local Modeling APPENDIX G 3D FLAC Local Modeling This appendix provides additional details regarding local models used to evaluate tailwall interlock pullout and wall defects within the OCSP® system. The intent of these evaluations was to determine: Whether the knuckle used in the OCSP® tailwall contributed significant extra pullout capacity to the tailwall as load developed on the facewall, and Whether defects within the existing OCSP® system represented significant additional risk to local stability during seismic loading. Results of these analyses are summarized in Section 7.5 of the suitability study report. In the following two sections, additional details regarding the tailwall interlock pullout and local defect models and the results of each analysis are provided. These results were used to form the conclusions given in Section 7.5 of the suitability study report. G.1 Tailwall Interlock Pullout Model The study of tailwall pullout was performed to address uncertainties in the PND design approach for interlock pullout resistance. The PND pull‐out resistance information was considered proprietary by PND (2008), and therefore assumptions had to be made by CH2M HILL during the suitability study regarding the amount of reaction that could be developed by each tailwall. It is CH2M HILL’s understanding from discussions with the Port of Anchorage (POA) and United States Army Corps of Engineers (USACE) that PND had conducted numerical analyses to show that the knuckles along the tailwall resulted in additional pullout capacity, relative to what would be developed from interface friction of backfill soil on flat steel sheet piles. Available information from PND also suggested that field tests have validated this additional reaction. Realizing this, a local numerical model was developed to investigate whether higher resistance would develop by pulling a knuckle through the backfill. The numerical modeling was conducted using FLAC3D to simulate a physical model test where the knuckle of the tailwall is pulled through granular backfill. The following subsections describe the geometry used in the model, steel and soil properties assigned to the model, the interface representation, boundary and initial conditions for the model, and the imposed loading. The final subsection summarizes results from the modeling effort. G.1.1 Geometry The FLAC3D model used to investigate the influence of interlocks on pullout resistance is based on pulling connected sheet piles from a box containing soil. Specifically, the model represents a series of connected sheet pile elements 1‐inch high that is pulled from a 1‐inch‐thick box containing soil. The numerical model allows the sheet piles to pass through openings in opposing walls in the box. The box sides facing the sheet pile elements in the model are open so that the upper and lower parts of the box can be filled with soil. The model has the ability to apply a uniform normal stress to the surface of the soil on the open sides of the box through a device similar to an inflatable diaphragm, which is held against the two sides of the sheet pile elements. The sheet pile geometry used in the models is based on PS31 sheet pile section geometry obtained from a CAD drawing from the web site of L.B. Foster, a sheet pile supplier. Figure G‐1 illustrates the sheet pile section geometry used in the pullout models. Figure G‐2 shows a section view of the test device when configured to contain one full sheet pile and two half‐sheets. Three different sheet pile‐interlock configurations were investigated: Two sheet pile halves One full sheet pile joined with two sheet pile halves Two full sheet piles with each one connected to the other and a sheet pile half ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX G-1 COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING Figures G‐3, G‐2, and G‐4 illustrate the three configurations, respectively. These configurations are subsequently referred to by the number of interlocks contained in the test specimen; that is, one, two, and three. Because the geometry in the pullout test model contains curves and small details, an automatic mesh generation software program was used to create unstructured finite element meshes of the model that could be imported into FLAC3D. Creating the meshes in FLAC3D would have taken considerable effort using FLAC3D’s built‐in structured meshing tools. Figure G‐5 illustrates a mesh of approximately 350,000 tetrahedron elements and 70,000 nodes created for the two interlock models. Element sizes were graded from small sizes at the sheet pile to larger sizes at the soil boundaries opposite the sheet pile elements. Because the depth of the upper and lower soil boxes in the model is a boundary condition that can influence the test results, it was also varied for the three sheet pile configuration models in order to investigate its effect on pullout resistance. A second set of models was created to increase the depth of the upper and lower soil boxes by 4 inches each. Table G‐1 summarizes the geometry of the pullout models. The models were run in FLAC3D’s large‐strain mode, which means that element geometry is updated at each calculation step to be consistent with the accumulated displacements. In addition, stress corrections were made for element rotations. Large‐strain mode was used because the pullout displacements in the test are large compared to the finite element sizes near the sheet pile. TABLE G‐1 Summary of Pullout Model Geometry Height, h inches Number of Tetrahedron Elements Number of Length, l Thickness, t Shallow Soil Deep Soil box Interlocks inches inches Box Box Shallow Soil Box Deep Soil Box One 20.7 1 14.9 22.9 240,306 258,736 Two 40.5 1 14.9 22.9 356,882 531,299 Three 60.3 1 14.9 22.9 606,697 656,226 G.1.2 Material Models There are multiple soil models in FLAC3D worth consideration for use in the pullout test model. However, in modeling it is generally best to begin with the simple and add complexity incrementally. This facilitates developing an understanding of the influence of particular factors. Beginning with a complex model of a system can hinder the modeling effort because it tends to flood the modeler with too much new information. Therefore, the Mohr‐ Coulomb (MC) elastic‐plastic model was used to represent the material response of the soil. The model parameters used are listed in Table G‐2. TABLE G‐2 Material Parameters for Pullout Test Physical Entity Material Model Parameters Soil Mohr‐Coulomb G=1540 psi K=3330 psi =32° (cid:31)=0° Steel Rigid N/A Soil/Steel Interface Mohr‐Coulomb k =2.5×105 lbf/in3 k=2.5×106 lbf/in3 (cid:31)=12° n s Although using steel for the sheet pile elements is the obvious modeling choice for a physical pullout test, the decision for numerical modeling requires more careful consideration. The choice of material model for the sheet piles has ramifications in numerical modeling, particularly with FLAC3D, which uses an explicit solution algorithm. G-2 ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING Using actual steel elastic material parameters for the sheet pile elements in the FLAC3D model would result in dramatically increased run‐times because of the large stiffness contrast between the soil and the steel. Models with large stiffness contrasts in general converge much slower than models with modest stiffness contrasts. There is not a precise definition of “modest stiffness contrast,” but generally it would involve ratios of maximum to minimum stiffness of 10 or less. The ratio of the elastic stiffness of steel to soil is much greater than this; therefore, because the steel is much stiffer than the soil, the sheet pile was modeled as a rigid material. This is a reasonable choice because the deformations of the steel sample under the test loads are several orders of magnitude less than the soil deformations. G.1.3 Interface Model The interface between the soil and the steel sheet pile also requires consideration. One modeling choice is to connect the soil elements directly to the steel sheet pile surface. This corresponds to an assumption of a no‐slip condition between the steel and soil, and experience indicates that slip does occur when soil is against a structural steel surface. In design, this interface is generally assigned a friction factor corresponding to a friction angle that is one‐half to two‐thirds of the soil. An interface friction angle of 12 degrees, which is a little less than one‐half of the soil friction angle for the backfill material, was used for most of the model tests. However, a few models were run with different values to study the effect of this parameter on the model results. Normal and shear stiffness values are also required in the FLAC3D interface material model. Unlike material parameters for solid constitutive models, the selection of interface stiffness values is more about achieving a computational goal; for example, limiting penetration across the interface. The normal stiffness value was set to about 10 times the smallest apparent normal stiffness of the adjoining soil elements in accordance with recommendations in the FLAC3D User’s Guide. The shear stiffness value was set to one‐tenth of the normal stiffness because the interface should be more compliant in shear. One issue was identified during modeling regarding the soil‐steel interface. During initial model development, interface elements were attached to the steel sheet pile on the element faces adjoining the soil. In FLAC3D interface elements are said to be one‐sided, which means that they work by detecting penetration of neighboring elements into the element faces that they are attached to. Figure G‐6 illustrates a zoomed view of the deformations at an interlock joint. It is apparent that some of the soil elements are penetrating quite far through the interface. Examination of the results indicates that this penetration is simply not detected by the interface elements. This issue seemed to be limited to areas of curvature on the interface and sharp corners. Based on commentary in the FLAC3D User’s Guide, a second opposing interface was added to the faces of the soil elements adjoining the steel elements. This effectively solved the issue, as illustrated in Figure G‐7, with the exception of some spurious results at the sharp interior corner in the interlock connection. However, because the elements involved are small, the forces involved are small, and the pullout force versus displacement relationship became much smoother after implementing the opposing interface elements, this approach was adopted. G.1.4 Boundary and Initial Conditions Boundary conditions are typically an important consideration in numerical modeling. One aspect of boundary conditions is the size of the test specimen, which was previously discussed under the heading Geometry. Also, the modeling approach taken was to view the pullout test as an adaptation of a direct shear test used in soil mechanics. Therefore, the sides of the soil box transverse to the sheet pile sample were modeled as being held by frictionless, rigid walls similar to a direct shear box. A confining stress was imposed on the soil by applying a uniform normal stress to the soil surface of the upper and lower soil boxes. These boundary conditions are illustrated in Figure G‐8. In addition, the sheet pile and soil are sandwiched between two parallel frictionless rigid surfaces, so no out‐of‐plane movements occur. In other words, the pullout test takes place under plane strain conditions. A normal stress of 10 psi was applied to the upper and lower soil boxes for all model tests; that is, =10 psi. The yy sand was initialized to at‐rest conditions with the normal stress on the two orthogonal planes set to 5 psi; that is, ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX G-3 COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING =5 psi. These conditions roughly correspond to stress conditions 15 feet below the ground surface if the xx zz groundwater table is at least 10 feet below the ground surface. Because the sheet pile sample is pulled out through an opening in the left‐hand wall of the device, it is difficult to envision any practical variation of the boundary condition for the left wall of the pullout device box other than to increase the distance from the nearby interlock joint. On the other hand, variations on the boundary conditions on the upper and lower soil surfaces and the right‐hand side of the device are more feasible. For instance, a confining stress could be applied through rigid platens resting on the soil. Similarly, a uniform normal stress can be applied to the right‐hand side of the soil boxes. Changing the right‐hand soil box boundary condition to an applied normal stress from a displacement condition (that is, smooth frictionless rigid wall) is easy to implement. On the other hand, modeling the confining load as being applied through rigid platens is more difficult to implement. Therefore, several models were run with the right side boundary condition changed to an applied normal stress of 5 psi. This was also done because the model results with the rigid right side boundary condition indicated loads were being transmitted to the right‐hand soil box side during the test; that is, the reaction on the right wall of the soil box decreased during the pullout test. G.1.5 Loading The pullout test was conducted by smoothly accelerating the steel sheet pile elements to a constant velocity and monitoring the reaction forces on the mesh. Through several trials it was found that a constant velocity of 2×10‐6 inches/step kept the unbalanced force ratio generally at or below 1×10‐4 during the test. Because FLAC3D uses an explicit solution algorithm, it is important to keep the unbalanced force ratio to a small value when material models with plasticity are present in the model as in this case. Otherwise, the transient stress waves transmitted through the model during solution could cause spurious solutions to develop. The acceleration of the sheet pile to the constant test velocity was controlled by the following interpolation function based on the trigonometric sine function: (cid:1874) (cid:3404) (cid:3049)(cid:3278)(cid:3290)(cid:3289)(cid:3294)(cid:3295)(cid:3276)(cid:3289)(cid:3295)(cid:3428)(cid:1871)(cid:1861)(cid:1866)(cid:3436)(cid:3420)(cid:3046)(cid:3047)(cid:3032)(cid:3043)(cid:2879)(cid:3046)(cid:3047)(cid:3032)(cid:3043)(cid:3116)(cid:3424)∙(cid:2024)(cid:3398)(cid:3095)(cid:3440)(cid:3397)1(cid:3432) (cid:2870) (cid:3015)(cid:3294)(cid:3295)(cid:3280)(cid:3291)(cid:3294) (cid:2870) where: step – is the current FLAC3D calculation step number step – is the calculation step number acceleration begins at 0 N – is the user‐specified number of steps to accelerate from 0 to v over step constant A similar function based on the cosine function is used to decelerate the sheet pile elements to zero velocity; that is, stop pulling. The models were generally set to pull the sheet pile out in 0.1‐inch increments and test for equilibrium at the stop points. G.1.6 Results One motivation for developing one, two, and three interlock pullout models was to examine the influence of sample length on pullout resistance. Another goal was to examine the influence of the boundary condition at the left side of the soil box; that is, a rigid frictionless wall. Figure G‐9 illustrates the results from the one, two, and three interlock models with a 12‐degree soil‐to‐steel interface friction angle. Several things are apparent from Figure G‐9. First, the ultimate pullout resistance of the single interlock model is between the theoretical ultimate pullout values of a rectangular bar having the same length and width as the sheet pile samples for interface friction angles of 12 degrees and 32 degrees. The theoretical pullout resistance of a rectangular bar is the product of the surface area of the bar times the normal stress, 10 psi, times the tangent of the interface friction angle times 2, because there are two sides in contact with the sand. The value of 12 degrees corresponds to the actual interface element shear strength, and the value of 32 degrees corresponds to the soil friction angle. Second, the increase in ultimate pullout resistance from the one interlock model to the two G-4 ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING interlock model is slightly smaller than a value equal to the increase in model length (one sheet pile width) times the friction force associated with 32 degrees. Third, the increase in ultimate pullout resistance from the two interlock model to the three interlock model is very small. Figure G‐10, which shows the displacement vectors at an approximate test displacement of 0.2‐inch, suggests why the increase in pullout resistance between the two and three interlock models is so small. It is apparent that a significant amount of soil is riding along with the third interlock. Figure G‐11 illustrates pullout results from the single interlock model for various values of the soil‐to‐steel interface friction angle. Although the interface friction angle influences the results, it is readily apparent that it is not a controlling factor. Specifically, the ultimate pullout resistance with an interface friction angle of 6 degrees is much larger than the theoretical value of a rectangular bar’s pullout resistance with the same interface friction. The increase in pullout resistance above the theoretical pullout resistance with low interface friction angles is a result of the interlocks engaging soil. Similarly, although the ultimate pullout resistance with an interface friction angle of 30 degrees is appreciably larger than with 6 degrees, it is still less than the theoretical pullout resistance on a rectangular bar with an interface friction angle of 32 degrees. The results from several pullout tests with the two‐interlock model are shown Figure G‐12. The models presented represent several different boundary condition cases, including shallow and deep soil boxes and using a stress boundary rather than displacement boundary on the right side of the soil box. Figure G‐13 presents similar results obtained with the one‐interlock model. As with the two‐interlock model, the one interlock model results show that although the boundary conditions influence the initial portion of the pullout displacement curve, the ultimate pullout resistance is essentially the same for the models. Figure G‐14 shows a deformed mesh and displacement contours from a two interlock model with the stress boundary condition on the right side of a deep soil box. The displacements, which are magnified by a factor of 10 in the figure, show the soil piling up against the left side of the model, which is fixed against displacement. This illustration highlights the significant role that the left side boundary condition has on the model test results. The majority of the reaction on the soil mass to the applied pullout load occurs on this boundary for cases in which a displacement boundary condition is used on the right side of the soil box. All of the reaction occurs on this boundary for the cases in which a stress boundary condition is used on the right side of the soil box. This observation, in conjunction with the effect of soil box depth on results, as illustrated in Figures G‐12 and G‐13, suggests that the soil box depth should be greater than the sheet pile sample length. G.2 Local Defect Model The second local model was developed to determine the potential effects of existing defects in the OCSP® facewall and tailwall. As discussed in Section 8 of this suitability study report, a number of construction defects occurred during installation of the sheet piles. These defects involved the sheets coming out of interlock during installation, caused by either driving on rock or by the effects of lateral dike loads on the sheets during driving. Gaps were identified in the facewall by divers during underwater inspections, and there were concerns that similar conditions could exist along the tailwall. These defects were identified as a potential source of further “unzipping” as the sheet piles reacted to additional loads from final dredging and during a seismic event. At the facewall, the unzipping would likely result in more loss of fill behind the wall, and if the amount of unzipping was large, could result in failure of the cell. For the tailwall unzipping could result in more load being transferred from the facewall to shorter intact sections of the tailwall, causing greater potential for complete unzipping as stresses progressively became higher. The following subsections discuss the geometry considered in the models, the method of modeling the OCSP® system, steel and soil properties used in the model, and the results of the evaluation. As part of this discussion, the differences between the local model used to evaluate these defects and the global model described in Sections 7.2 through 7.4 of the suitability study report are also summarized. G.2.1 Geometry It is possible to identify many different scenarios of a facewall or tailwall defect based on differences in OCSP® geometry, subsurface conditions, and so on for the project. It is not practical to attempt to model all of the ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX G-5 COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING different scenarios because this could easily lead to many hundreds of model instances and a modeling effort that might take a year or more to complete. Therefore, the approach taken was to use modeling to analyze a basic problem and develop a better fundamental understanding of system performance. This approach provides a better basis for making engineering judgments when addressing the issues. Two basic models with different geometry were developed. The first model serves as a reference or baseline model for the second model. The first model represents an infinitely long OCSP® wall at the maximum design section without any defects. Because of the symmetry of an infinitely long wall, this model is limited to two half cells and one tailwall, the same as the previously presented FLAC3D primary model. This first model is referred to by the nomenclature “2h” as shorthand for “two half‐cell” in this section. Figure G‐15 illustrates the FLAC3D mesh for the 2h model. The 2h model contains 65,400 zones (soil elements) and 5,656 structural elements. The second model is used to model isolated defects in a facewall and tailwall. Because the defects are intended to be isolated, the model must encompass a greater reach of the wall than the reference model; however, a large model means increased model run time and memory requirements. Therefore, the need to model a larger reach of wall must be balanced with the need to obtain a model that does not exceed the memory capacity of the software and modeling platform and runs in a reasonable period of time. In this case, this balance meant limiting the second model to two cell widths; that is, a width of 55 feet. The second model is arranged so that one complete cell and two half cells with two tailwalls are modeled. The nomenclature “1w+2h” is used as shorthand for “one whole cell and two half‐cells” to identify the second model in this section. Figure G‐16 illustrates the FLAC3D mesh for the 1w+2h model, which contains 130,800 zones (soil elements) and 11,344 structural elements. Because the focus of the local defect modeling effort is on OCSP® stresses, emphasis was placed on attaining a high resolution of the numerical model mesh in the vicinity of the OCSP® face and tailwalls. There are seventeen PS31 steel sheet piles in the facewall of each of the ±27‐foot‐wide cells. The aspect ratio of elements influences the accuracy of the numerical solution and a hexahedron element with height to width and width to depth ratio of 1 provides the most accurate solution. Therefore, in order to maintain an aspect ratio as close to 1 as possible, an element width slightly less than one PS31 sheet was used in order to divide each whole cell into 16 elements and each half cell into 8 elements. Maintaining this mesh resolution throughout the model required many elements and thus quickly escalated run time and memory requirements. Therefore, the mesh size away from the wall was increased. FLAC3D allows unconforming meshes (that is, meshes with different element sizes) to be attached together along common planar boundaries. The gridpoints (nodes) of the finer mesh are slaved to the displacements of the coarser mesh along the common boundary. This slaving works best when the coarse and fine meshes are related by an integer multiple of elements along each boundary segment. In the 2h and 1w+2h models, coarse meshes with element sizes double that of the finer mesh around the OCSP® bulkhead were attached between the inner finer mesh and the model boundaries. G.2.2 OCSP® System The facewall and tailwall were modeled using three‐noded, flat membrane type plate structural shell elements. The membrane elements have four degrees of freedom (DOF) at each node, three of which are translational and one is an in‐plane (also referred to as drilling) rotation, and use a stiffness formulation known as constant strain triangle hybrid stress. The element formulations in FLAC3D allow isotropic, orthotropic, or anisotropic elastic material response. As previously described in this section, structural elements in FLAC3D interact with the three‐dimensional continuum mesh via linkages at the structural element nodes to the zones. The linkages are created automatically for various classes of structural elements. The tailwall is modeled with the geogrid class of elements for which the structural nodes are slaved to the movements of the continuum mesh in the direction normal to the face of the geogrid elements; however, in the transverse or shearing direction, a coupling shear spring allows relative movement between the geogrid elements and the continuum mesh. The shear stresses that develop in the coupling spring are limited by a Coulomb‐type strength law with cohesion and friction. The effective normal G-6 ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING stresses in the adjacent three‐dimensional soil zones are used with the assigned link friction angle and cohesion to calculate the frictional strength component of each spring. The shear spring parameters for the tailwall in the local defect models are given in Table G‐3. TABLE G‐3 Structural Element Linkage Parameters for Local Defect Models Parameter Name Value k Normal stiffness 2×107 lbf/ft3 n k Shear stiffness 2×105 lbf/ft3 s c Cohesion 0 δ Friction angle 30° c Residual cohesion 0 res N Tension cutoff strength 0 cut The normal stiffness of the linkage springs was selected so that the normal movements between the soil zones and the structural elements is small and insignificant relative to the soil movements, but not so stiff as to create a large stiffness contrast in the model that would adversely affect convergence. The shear stiffness was set to a fraction of the normal stiffness in order to have more compliance in the shear direction, similar to a real interface. A friction angle of 30 degrees was used for shear strength of the interface because the pullout model results described in Section G.1 suggest it is reasonable to use a friction angle slightly less than the granular backfill’s 36‐ degree friction angle. The facewall is modeled using the liner element class. For the tailwall, the structural element is a 3‐noded membrane element. The embedded liner option is used, which means that each liner element can interact with soil elements on each face of the element. The soil elements on either side of an embedded liner element are not joined to the elements on the opposite side. On each side of the liner, normal and shear springs couple the liner to the adjoining soil elements. The shear spring formulation for the liner element is the same as for the geogrid element except the normal stress for the Coulomb‐type shear strength equation comes from the normal spring response. In FLAC3D, structures composed of different structural element classes do not interact with each other unless linkages are made between the structural nodes. In the local defect model, linkages are made at the nodes of the facewall and tailwall that lie on the vertical line corresponding to the wye connection in the physical OCSP® system. In the local defect model, the linkages made are to rigidly slave the facewall nodes to the tailwall geogrid nodes at the wye location in the two transverse translational DOF in the horizontal plane. However, in the vertical direction the nodes are allowed to move independently. This corresponds to an assumption of a sheet pile interlock that has no slack and that is also perfectly smooth; that is, offers no frictional resistance to sliding. This condition is analogous to a sliding hinge. The alternative would have been to rigidly slave the two nodes in the vertical direction, but this was not done because this would correspond essentially to a fixed hinge. Because membrane type shell elements, which only utilize translation degrees of freedom and an in‐plane rotation, were used, the hinge action is insignificant to the model behavior. As often happens in numerical modeling, the behavior of the real system lies between the two bounding cases available in the numerical model. In these situations, a choice must be made to use one of the bounding cases or to create two models and judge where the real system response is relative to the two bounding cases. The discussion about the wye‐connection interlock is related to the issue of how to model all of the interlocks in the OCSP® system. An actual interlock connection between sheet piles is characterized by several important mechanical characteristics. First, the opposing “finger‐and‐thumb” of two sheet piles that interlock do not fit perfectly; therefore, some slack must be taken up before tension or compression can be transmitted across the ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX G-7 COPYRIGHT 2012 BY CH2M HILL APPENDIX G FLAC3D LOCAL MODELING connection. Also, because bearing surfaces in the interlock become larger at the points of contact when the interlock forces increase as a result of elastic and inelastic deformation at the contact, the force‐displacement relationship of the interlock stiffens as the force increases. The ability of the finger and thumb to bend also contributes to compliance of the interlock. The second important characteristic of a sheet pile interlock is for friction along the points of contact to resist shearing/sliding action on the interlock; that is, one sheet moving up or down relative to the other. The wye connection is the only interlock in the OCSP® system that is directly modeled with a mechanism in the local defect model. It is easy to look at a visualization of the FLAC3D meshes and misinterpret the columns of structural elements in the facewall and tailwall as representing individual sheet piles. However, this is not the case and the elements and nodes on the facewalls and tailwall represent a discretization of continuous steel plate structures. It would obviously be desirable to directly model all the interlocks, and some effort was expended on several concepts for how to do this using the built‐in link mechanisms augmented by some customizations with FLAC3D’s built‐in programming language. However, several obstacles were encountered that could not be overcome in the time available. Therefore, an indirect means of modeling the effect of interlocks on sheet pile structures was used. The indirect means of modeling interlocks is to use an orthotropic elastic material model instead of an isotropic linear elastic material model for the structural elements. There are two components to this indirect approach. The first is to use the material parameters to reduce the shear stiffness of the assemblage of structural elements comprising a tailwall or facewall to more closely resemble the shear stiffness of an assemblage of discrete sheet piles with interlocks of limited shear strength, instead of a continuous sheet of steel. The orthotropic material stiffness matrix was set equal to the isotropic material stiffness matrix with adjustments to two terms. The isotropic material stiffness matrix for steel (E=29×106 psi, ν=0.25) is: (cid:3006) (cid:3006) (cid:2021)(cid:4672) (cid:4673) 0 (cid:1741) (cid:1744) (cid:1855)(cid:2869)(cid:2869) (cid:1855)(cid:2869)(cid:2870) (cid:1855)(cid:2869)(cid:2871) (cid:2869)(cid:2879)(cid:3092)(cid:3118) (cid:2869)(cid:2879)(cid:3092)(cid:3118) 4.54 1.14 0 (cid:3429)(cid:1855)(cid:2869)(cid:2870) (cid:1855)(cid:2870)(cid:2870) (cid:1855)(cid:2870)(cid:2871)(cid:3433) (cid:3404) (cid:1742)(cid:1742)(cid:2021)(cid:4672) (cid:3006) (cid:4673) (cid:3006) 0 (cid:1745)(cid:1745)=(cid:3429)1.14 4.54 0 (cid:3433)(cid:3400)10(cid:2877)(cid:1868)(cid:1871)(cid:1858) (cid:2869)(cid:2879)(cid:3092)(cid:3118) (cid:2869)(cid:2879)(cid:3092)(cid:3118) (cid:1855)(cid:2869)(cid:2871) (cid:1855)(cid:2870)(cid:2871) (cid:1855)(cid:2871)(cid:2871) (cid:1742) (cid:3006) (cid:1745) 0 0 1.70 0 0 (cid:1743) (cid:2870)(cid:4666)(cid:2869)(cid:2878)(cid:3092)(cid:4667)(cid:1746) where: (cid:2026) (cid:1855) (cid:1855) (cid:1855) (cid:2013) (cid:3051) (cid:2869)(cid:2869) (cid:2869)(cid:2870) (cid:2869)(cid:2871) (cid:3051) (cid:4668)(cid:2026)(cid:4669) (cid:3404) (cid:4670)(cid:1831)(cid:4671)(cid:4668)(cid:2013)(cid:4669) (cid:3404) (cid:3421)(cid:2026)(cid:3052)(cid:3425) (cid:3404) (cid:3429)(cid:1855)(cid:2869)(cid:2870) (cid:1855)(cid:2870)(cid:2870) (cid:1855)(cid:2870)(cid:2871)(cid:3433)(cid:3421)(cid:2013)(cid:3052)(cid:3425) (cid:2028) (cid:1855) (cid:1855) (cid:1855) (cid:2011) (cid:3051)(cid:3052) (cid:2869)(cid:2871) (cid:2870)(cid:2871) (cid:2871)(cid:2871) (cid:3051)(cid:3052) The coupling term between the vertical and horizontal normal strains and stresses and the shear stiffness constant were each reduced by a factor of 0.01. Therefore, the orthotropic material stiffness matrix used is: 4.54 0.01 0 (cid:3429)0.01 4.54 0 (cid:3433)(cid:3400)10(cid:2877)(cid:1868)(cid:1871)(cid:1858) 0 0 0.017 The second component of using the material model of the tailwall and facewall is to reduce the stiffness matrix term relating lateral strain, (cid:2013) , to lateral stress, (cid:2013) . Previous investigators using finite element methods to model (cid:3051) (cid:3051) cellular cofferdam structures have determined that reducing the stiffness in the lateral direction by a factor of 1/100 to 1/30 yields results that are reasonably close to earth pressures and deformations observed when filling circular cofferdam cells. Because the goal of the local defect modeling is to understand the impact of defects, it is sufficient to examine differences in the models with and without defects. Therefore, the values from any one model are less significant than the difference in values with another model that is identical in all respects except for the presence of a defect. This means that factors such as interlock compliance, which would essentially equally affect the models, can be omitted, while recognizing that this will affect the total displacements and stresses more than the differentials between the models. Consequently, a reduction factor was not used for the defect models. This was judged to be a reasonable representation of the system after backfilling is complete when other loads such as dredging occur, because the slack would mostly be taken up during filling. G-8 ANC/APPG_LOCAL_MODELING_100PERCENTDRAFTFINAL_WITHOUT.DOCX COPYRIGHT 2012 BY CH2M HILL
Description: