NCHRP -68 FY 4 Rotational Limits for Elastomeric Bearings Final Report APPENDIX E John F. Stanton Charles W. Roeder Peter Mackenzie-Helnwein Department of Civil and Environmental Engineering Universit of Washington Seattle, WA 9895-7 - i -
YOUR NOTES - ii -
TABLE OF CONTENTS APPENDIX E FINITE ELEMENT ANALYSIS E- E. Introduction E- E. Method and Model definition E- E.. Method E- E.. Loading E- E.. Geometr and boundar conditions E- E..4 Materials E-5 E. Evaluation and Validation of Stiffness Coefficients E-7 E.. Axial Stiffness Coefficient B E-8 E... Strip-bearing with SF 6 E-8 E... Strip-bearing with SF 9 E- E... Strip-bearing with SF E- E...4 Summar for the axial stiffness coefficient B E-6 E.. Bending Stiffness Coefficient B r E-8 E... Strip-bearing with SF 9 Rotation at σ =.GS E-9 E... Strip-bearing with SF 9 Rotation at σ =.GS E- E... Strip-bearing with SF 9 Rotation at σ =.5GS E- E...4 Strip-bearing with SF 9 Rotation at σ =.85GS E-4 E.. Summar for the rotational stiffness coefficient B r E-4 E.4 Evaluation of Local Shear Strain for Design Purposes E-6 - iii -
E.4. Local Effects at the End of the Steel Shims E-6 E.4. Shear Strain Coefficient C zx for Axial Loading E- E.4.. Strip-bearing with SF 6 E- E.4.. Strip-bearing with SF 9 E- E.4.. Strip-bearing with SF E- E.4. Shear Strain Coefficient C rzx for Rotation Loading E- E.4.4 Effect of lift-off on the local shear strain E-5 E.4.5 Bearings with Bonded External Plates Uplift E-8 E.5 Model Error Significance of nonlinear effects and Superposition E-4 E.6 Conclusions E-54 LIST OF FIGURES Figure E- Geometr and Finite Element mesh for a strip bearing with SF 6...E-4 Figure E- Geometr and Finite Element mesh for a strip bearing with SF 9...E-4 Figure E- Geometr and Finite Element mesh for a strip bearing with SF...E-4 Figure E-4 B versus average axial strain for bearing with SF 6...E-9 Figure E-5 B versus average axial strain for bearing with nominal SF 6 and calculation based on theoretical SF 5.67...E- Figure E-6 B versus average axial strain for bearing with nominal SF 6 and calculation based on the average SF 5.8...E- Figure E-7 B versus average axial strain for bearing with SF 9...E- Figure E-8 B versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5...E- Figure E-9 B versus average axial strain for bearing with nominal SF 9 and calculation based on the average SF 8.75...E- - iv -
Figure E- Figure E- σ / GS versus average axial strain for bearing with SF...E-4 B versus average axial strain for bearing with SF...E-4 Figure E- B versus average axial strain for bearing with nominal SF and calculation based on theoretical SF...E-5 Figure E- B versus average axial strain for bearing with nominal SF and calculation based on the average SF.67....E-6 Figure E-4 Hourglass modes distorting the right side of a bearing with bonded external plates subjected to rotation onl...e-9 Figure E-5 B r versus average axial strain for bearing with SF 9 at P =...E- Figure E-6 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =...E- Figure E-7 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =...E- Figure E-8 B r versus average axial strain for bearing with SF 9 at P =. GS A..E- Figure E-9 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =. GS A...E- Figure E- B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =. GS A...E- Figure E- B r versus average axial strain for bearing with SF 9 at P =.5 GS A..E- Figure E- B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =.5 GS A...E- Figure E- B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =.5 GS A...E- Figure E-4 B r versus average axial strain for bearing with SF 9 at P =.89 GS A E-5 Figure E-5 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =.89 GS A...E-5 - v -
Figure E-6 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =.89 GS A...E-5 Figure E-7 Computed shear strain γ = Ε in (in/in) at σ =.5GS at the edge of a steel shim....e-7 Figure E-8 Computed shear strain γ = Ε in (in/in) at σ =.GS at the edge of a steel shim....e-8 Figure E-9 Computed hdrostatic stress σ in (Pa) at σ =.5GS at the edge of a steel shim....e-9 Figure E- Computed hdrostatic stress σ in (Pa) at σ =.GS at the edge of a steel shim....e- Figure E- Shear strain along the shim at different axial loads for a bearing with SF 9...E- Figure E- Figure E- Figure E-4 C versus σ / GS for bearing with SF 6...E- zx C versus σ / GS for bearing with SF 9...E- zx C versus σ / GS for bearing with SF...E- zx Figure E-5 Coefficients for bearing with SF 9...E-4 Figure E-6 Shear strain along a shim of a -laer bearing with SF 9: Axial load level σ =.5 GS and total rotation θ =. /./.5/ 5.%...E-6 Figure E-7 Shear strain along a shim of a -laer bearing with SF 9: Axial load level σ =.5 GS and total rotation θ =. /./.5/ 5.%...E-6 Figure E-8 Lift-off ratio versus rotation per laer for SF 9 bearing at various load levels...e-7 Figure E-9 Tensile hdrostatic stress in (Pa) for σ =. GS and θ =.6 rad/laer...e-9 Figure E-4 Tensile hdrostatic stress in (Pa) for σ =.6 GS and θ =.6 rad/laer...e-9 Figure E-4 Tensile hdrostatic stress in (Pa) for σ =.445 GS and θ =.6 rad/laer...e-4 - vi -
Figure E-4 Histor of hdrostatic stress in (psi) for σ =. GS (Increments 6-9)...E-4 Figure E-4 Histor of hdrostatic stress in (psi) forσ =.6 GS (Increments 6-9)E- 4 Figure E-44 Histor of hdrostatic stress in (psi) for σ =.445 GS (Increments 6-9)...E-4 Figure E-45 Collection of resulting shear strain (in/in) from different nonlinear analses of an elastomeric bearing with SF 9...E-44 Figure E-46 Fourth-order fit to simulation data for a SF 6 bearing with bonded external plates for shear strain at ¼ distance from the critical edge of the shim. Simulated data (lines/curves), fitted surface (left image), and contour lines for shear strain (in/in)....e-48 Figure E-47 Fourth-order fit to simulation data for a SF 9 bearing with bonded external plates for shear strain at ¼ distance from the critical edge of the shim. Simulated data (lines/curves), fitted surface (left image), and contour lines for shear strain (in/in)....e-49 Figure E-48 Fourth-order fit to simulation data for a SF 6 bearing without bonded external plates for shear strain at ¼ distance from the critical edge of the shim. Simulated data (lines/curves), fitted surface (left image), and contour lines for shear strain (in/in)...e-5 Figure E-49 Fourth-order fit to simulation data for a SF 9 bearing without bonded external plates for shear strain at ¼ distance from the critical edge of the shim. Simulated data (lines/curves), fitted surface (left image), and contour lines for shear strain (in/in)...e-5 Figure E-5 Iso-error plot for shear strain γ zx for a SF 6 bearing with bonded external plates based on a fourth-order fit. (Constructed for shear strain at ¼ distance from the critical edge of the shim.)...e-5 Figure E-5 Iso-error plot for shear strain γ zx for a SF 9 bearing with bonded external plates based on a fourth-order fit. (Constructed for shear strain at ¼ distance from the critical edge of the shim.)...e-5 Figure E-5 Iso-error plot for shear strain γ zx for a SF 6 bearing without bonded external plates based on a fourth-order fit. (Constructed for shear strain at ¼ distance from the critical edge of the shim.)...e-5 Figure E-5 Iso-error plot for shear strain γ zx for a SF 9 bearing without bonded external plates based on a fourth-order fit. (Constructed for shear strain at ¼ distance from the critical edge of the shim.)...e-5 - vii -
LIST OF TABLES Table E- Stiffness coefficients used b Gent & Meinecke (97) and Stanton & Lund (4) and their simplified equivalent in the proposed design equations...e-7 Table E- Model parameters for SF 6 bearing...e-9 Table E- Model parameters for SF 9 bearing...e- Table E-4 Model parameters for SF bearing...e- Table E-5 Comparison of stiffness coefficient B for modified definitions of S...E-7 Table E-6 Coefficients a ij for a bearing with SF 6 and with bonded external plates (creates uplift)...e-45 Table E-7 Coefficients a ij for a bearing with SF 9 and with bonded external plates (creates uplift)...e-46 Table E-8 Coefficients a ij for a bearing with SF 6 without bonded external plates (allows lift-off)...e-47 Table E-9 Coefficients a ij for a bearing with SF 9 without bonded external plates (allows lift-off)...e-47 - viii -
APPENDIX E Finite Element Analsis E. Introduction The performed project focused mainl on the experimental investigation of the rotational behavior of elastomeric bearings and the development of suitable design procedures. This requires the identification of characteristic engineering design parameters which () can be measured in the experiment (direct or indirect), and () are reasonabl simple to use for the routine design of bearings. One ke problem arising with experimental work is the selection of representative measures which can be tracked during testing. This restricts in-situ experimental observation to visible and measurable behavior on the surface of each tested bearing. Interior changes such as damage to the elastomer, permanent deformations in the steel shims and/or the elastomer, or damage to the interface between these components can onl be observed post mortem, i.e. after the test program has finished. Local high strains in the interior of the bearing cannot be measured at all since a measurement would require modifications of the surface cover close to the end of the steel shims. This in turn would induce initial damage to exactl that portion of the bearing where initial debonding shall be used to identif the onset of structural damage to the bearing. These limitations of experimental procedures ma result in observations which are often hard to interpret, and can void their relevance as a damage measure. Finite element analsis (FEA) provides a tool to simulate structural behavior and assess internal states of deformation and stress. Using FEA enables us to correlate experimentall observed or controlled measures with local (internal) quantities. This benefit, however, is limited to the simulation of a well defined structure made of idealized homogeneous material. The level of accurac depends on the tpe of material and the qualit of available material properties. Despite this limitation, FEA is used to identif the relationship between external loads or applied displacements and rotations, and the engineering design parameters used in a design procedure. These relationships are then used to evaluate and justif simplified design equations. The performed structural testing and evaluation of the experimental data identified the local shear strain γ zx in the elastomer at (or near) the ends of the reinforcing steel shims as suitable engineering design parameter. Gent & Meinecke (97) presented a simplified linear analsis for incompressible elastometric bearings. Stanton & Lund (4) extended this theor for slightl compressible elastomers. Both formulations provide simple relations between axial force and average axial strain, as well as between moment and rotation. The shape of the bearings enters these relations solel through the shape factor S of the bearing. The also provide a convenient correlation between those global deformation measures and the local shear strain γ inside the bearing. The linearit of zx, max the equations b Stanton & Lund (4) provides an excellent basis for a design procedure. Their approach, however, utilizes superposition of axial and rotational loading and its validit has to be proven. - E- -
This chapter presents a numerical analsis b which the following hpotheses will be proven to hold for common bearings:. The stiffness coefficients predicted b the linear theor of bearings b Stanton & Lund are in good agreement with a nonlinear FEA.. The local shear strain predicted b the linear theor of bearings b Stanton & Lund are in good agreement with a nonlinear FEA.. Superposition of axial and rotational effects provides a reasonabl accurate representation of the nonlinear FEA. 4. Internal rupture due to excessive tensile hdrostatic stress can onl occur in bearings with bonded external plates and ver low axial load intensities. Proving these three hpotheses is of crucial importance for the justification for using the simple linear analsis b Stanton & Lund for design purposes. E. Method and Model definition E.. Method All nonlinear analsis presented in this chapter was performed using the multi-purpose Finite Element program MSC.Marcr b MSC.software. It provides a wide variet of specialized elements and material formulations. Through a long histor of cooperation between MSC.software and the tire industr, MSC.marcr offers a good selection of material models for rubber and rubber-like materials. Test simulations revealed that the commonl known problems of large deformation simulations of elastomers, i.e. extensive mesh distortion and the potential loss of element stabilit (visible as hourglass modes) impose a serious limitation on the numerical analsis. Local mesh distortions are alwas observed near the end of the reinforcing steel shims. As loading progresses, mesh distortion inevitabl results in inversion of elements and subsequent failure of the analsis. This limits the maximum axial compression and rotation which can be analzed to average stresses of σ = GS and rotations of.8-. rad/laer. Hourglass modes were observed in areas of a bearing where hdrostatic tension occurs. The performed analses showed reliable results for hdrostatic tensile stresses (= positive values of mean normal stress) up to σ = G = E, with shear modulus G and Young s modulus E. This is approximatel the magnitude described b Gent & Lindle (958) for the generation of internal rupture. Hence, the FEA model can be used to assess critical hdrostatic tension in the bearings. - E- -
All analses were performed on strip bearings using a large deformation plane strain analsis in a Lagrange setting. Three-dimensional analses could not be performed at the necessar level of refinement due to numerical instabilities of the nearl incompressible element formulation at high hdrostatic stress. E.. Loading One major consideration of creating the model was how to realisticall portra the loading conditions. In all of the experimental tests the bearing was placed between two metal plates to ensure uniform loading. To simulate the experimental conditions the loading plates were represented using rigid surfaces and frictional contact between these surfaces and the bearings. The motion of the rigid loading surface can be controlled b defining the position (or velocit) of a reference point on the surface and an angle (or angular velocit) to identif its orientation. This requires the loading to be displacement controlled rather than, as desired, load controlled. The loading of the bearing was composed of slow axial compression to various levels of average axial strain ε, followed b rigid bod rotation of the loading surface. Forces and moments on the loading surface, as well as local strain and stress measures were recorded for each load histor. In order to discuss the method used to define the model, there are three main aspects that will be considered. The are the geometr of the model and the boundar conditions, the material models, and the loading. E.. Geometr and boundar conditions Bearings of various shape factor were considered in the numerical analsis. The most representative bearings were those with a cross section similar to the ones tested in the experimental program. An identical cross section as the tested 9 in x in bearings with SF 6 ields a strip bearing with SF 9. (This is an effect of the different aspect ratio.) A model with SF 6 was generated b increasing the laer thickness from.5 in to.75 in. For validation purposes, a SF bearing of the same tpe was obtained b reducing the laer thickness to.75 in. All two-dimensional finite element models possess two smmetr planes. Due to the applied rotation, the loading patterns possess onl one smmetr plane. The model utilizes the horizontal smmetr plane which both the bearing structure and the loading pattern have in common. This plane goes through the center of the bearing. Smmetr boundar conditions were applied on the smmetr plane. Horizontal stabilit of the model was achieved through a ver soft horizontal spring with vanishing stiffness once the loading surfaces are moved into contact with the bearing. This is necessar since the bearing will be held in place b friction on the loading surfaces but will move horizontall in the smmetr plane when rotation is applied at the loading surface. This would leave the bearing without horizontal support prior to contact with the loading surface except for the artificial soft spring. - E- -
Figure E-, Figure E-, and Figure E- show the geometr and the tpicall used Finite Element mesh for bearings with SF 6, SF 9, and SF, respectivel. (Onl top half modeled; not to scale.) The indicated reference point is used for shear strain analsis in sections E.4 and E.5..5 in. 8.5 in..5 in.. in..75 in.. in..75 in. Reference Point A Figure E- Geometr and Finite Element mesh for a strip bearing with SF 6.5 in. 8.5 in..5 in..5 in.. in..5 in.. in..5 in. Reference Point A Figure E- Geometr and Finite Element mesh for a strip bearing with SF 9.5 in. 8.5 in..5 in. Reference Point A.5 in.. in..75 in.. in..75 in.. in..875 in. Figure E- Geometr and Finite Element mesh for a strip bearing with SF Bearings with rigid or semi-rigid end-plates are modeled using the same mesh and loading but changing the interface properties from frictional contact to glue. This permanentl attaches the bearing surface to the rigid loading surface and allows for identical load histories on bearings with and without rigid end-plates. - E-4 -
Elements representing elastomer were modeled using quadrilateral bi-linear elements with constant dilation, also known as Herrmann formulation [MSC.marc Manual A & B ()] (element tpe 8 in MSC.Marcr). This formulation was designed for the analsis of incompressible and nearl incompressible material. Elements representing the steels shims were modeled using quadrilaterals bi-linear shape functions with full integration (element tpe in MSC.Marcr). Tests on alternative integration schemes and mixed formulations produced no noticeable difference in the solution. While the experimental data was collected and analzed entirel using English sstem units, it was more convenient using metric units in the numerical analsis. The units used to run the analtical model in MSC.Marcr were meters (m) for length and Newton (N) for force, resulting in Pascal (Pa) for stress. These are the units on images and diagrams produced directl b MSC.marcr. In order to analze the data and to present comparisons with the linear theor b Stanton & Lund, the units were converted to the customar English sstem. That is, inches (in) were used for length, kilo-pounds (kips) for force, and kips per square inch (ksi) for stress. E..4 Materials The elastomer material was modeled as nonlinear elastic, nearl incompressible material. MSC.marcr provides the Ogden model, a generalized Moone model, and the extension to the generalized Moone model known as Gent s model [MSC.marcr, manual A]. These models require between 8 and parameters. Precisel speaking, the general Ogden model can be extended to an even number of parameters but MSC.marcr limits it for practical reasons [Ogden (984)]. Bearing manufacturer on the other hand have to test their elastomer material and report Shore A hardness, elongation at failure and nominal stress at failure as the onl parameters relevant for the D characterization of the elastic material. This presents a challenge and leaves space for variabilit on the manufacturer s side, and non-uniqueness of material parameters on the modeling part. Following a carefull justified suggestion b the project advisor group, the simplifing assumptions b Yeoh (99) were adopted. Shore A hardness was used to estimate the shear modulus G. This parameter was later confirmed b shear tests performed on elastomer samples extracted from actual bearings. The bulk modulus K was obtained and later verified through a similar process. Parameters for Yeoh s model were obtained as follows. First, a master tension test result was selected from the literature. This master curve is scaled linearl along the strain (or stretch) axis to match the ultimate elongation at break known from the manufacturer s testing. Then the stress axis is stretched b a linearl varing scaling function such that (a) the ultimate stress corresponds with the manufacturer s test, and (b) that the initial slope of the scaled curve (=uniaxial stiffness E ) matches Young s modulus E = G as identified from the Shore A hardness. This procedure makes sense since most materials used for elastomeric bearings have a ver - E-5 -
similar shape of the uniaxial tension curve and differ mainl in few characteristic stiffness and strength values. Second, Yeoh s model can be represented as a sensible subset of the generalized Moone model and as such can be easil implemented in MSC.Marcr. The work function for Yeoh s model is W ( E ) = C I, (E-) ( I ) + C ( I ) + C ( ) where E is the Green-Lagrange strain tensor, I = λ + λ + λ = tr C is the first invariant of the right Cauch-Green strain tensor C = + E, and λ i, i =,, are the principal stretches. The nd Piola-Kirchhoff stress tensor follows from (E-) as I [ C + C ( I ) + C ( I ) ] + p W E S = C E E ( ) + pc = with the hdrostatic stress p to be determined from equilibrium equations., (E-) Other stress measures follow from standard transformation rules. (See, e.g., Malvern (969), Gurtin (98), Ogden (984), Marsden & Hughes (98) for details.) Equation (E-) can be specialized for uniaxial tension and leads to the engineering stress with σ * * [ ] + C ( I ) + C ( I ) λ λ Force = C. (E-) Reference Area = * I = λ +. (E-4) λ Equation (E-) is used to identif the parameters C, C, C, using a least square fit of the scaled material test. The initial stiffness E used to scale the reference material test is obtained as the initial slope of the stress-strain curve. It follows from (E-) as dσ E = = 6 dλ λ = C. (E-5) The following parameters were commonl used for the numerical analses: shear modulus G = psi (.689 MPa), Bulk Modulus K = 4 ksi = 4, psi (758 MPa), C = 44,47 Pa (49.96 psi), C = -6,6 Pa (.9555 psi), and - E-6 -
C = 9.9 Pa (.464 psi). The steel shims in the bearings were A Gr 6 steel. (Specification provided b an undisclosed manufacturer.) The material was modeled using isotropic, linear elasticperfectl plastic J -plasticit (D/D von Mises material). The material properties used in the FEA were Young s Modulus E = GPa (9, ksi), Poisson s ratio ν =., and a ield stress of σ = 5 MPa (4. ksi). Y E. Evaluation and Validation of Stiffness Coefficients Gent & Meinecke (97) demonstrated that axial stiffness and rotational stiffness of bearings can be expressed in terms of shear modulus G and shape factor S. Introducing four coefficients A, B, Ar, Br ields the following relations for the equivalent axial stiffness E = G( A B S ) (E-6) + and the effective bending stiffness E I = G( A + B S I. (E-7) r r r ) Stanton & Lund (4) extended the formulation for slightl compressible material and proved that b properl adjusting the stiffness coefficients, the general form of equations 4 (E-6) and (E-7) remains valid. The also showed that A ax = A r = for all strip bearings. The subscripts introduced b Gent & Meinecke (97) and Stanton & Lund (4) are used throughout Appendix E. However, for the sake of simplifing the notation, a simpler and more compact notation will be introduced and used in subsequent Appendices concerning design equations. For the reader s convenience, a list of equivalent smbols is given in Table E-. Table E- Stiffness coefficients used b Gent & Meinecke (97) and Stanton & Lund (4) and their simplified equivalent in the proposed design equations Notation b Gent & Meinecke (97) and Stanton & Lund (4) A, A, r B B r Simplified (compacted) notation used in the proposed design equations A, B a A, C zx, C rzx C a, r B a r C r - E-7 -
The aim of this section is to back-calculate these coefficients based on numerical results from nonlinear FEA. This will provide insight on both the significance of nonlinearit over the common load range for elastomeric bearings and on the potential model error of the linear theor b Stanton & Lund (4). E.. Axial Stiffness Coefficient B The bearing analsis is based on an infinite strip bearing. All simulations presented in this section were performed for axial compression onl. Numerical instabilities limited the analses to an average vertical stresses of σ = GS. Obtained numerical data represents the actual range of loading in real bearing applications, but lies below the experimentall applied load levels. The numerical analsis provides force, displacements, local strains, and stresses. The displacement data is used to compute the average axial strain ε δ top δ bottom =, (E-8) t where δ top and δ bottom are the vertical displacements of the shim on top and bottom, respectivel, of the elastomer laer of thickness t. The average axial strain ε is positive if the bearing is compressed and negative if elongated. Force p as force per unit width of the bearing is then used to compute the equivalent stiffness coefficient B p = A S ELε (E-9) with L as the length of the bearing (= 9 in), A = 4 /, and E = G G. (E-) + G / K The error for using E = G will be below % for all common elastomers, and henceforth we will use the relation E = G instead of (E-). E... Strip-bearing with SF 6 The bearing analsis is based on an infinite strip bearing as shown in Figure E- and dimensions as given in Table E-. The analsis was performed for axial compression onl. Numerical instabilities limited the analsis to an axial load of P =.kips/in (or an equivalent of P = 464 kips on a 9 in x in bearing).this represents an average stress of σ =.98GS. a - E-8 -
Table E- Model parameters for SF 6 bearing K 4, psi (,76. MPa) G psi (.69 MPa) E 99.98 psi (.68 MPa) h.laer.75 in n.laer L 9 in Figure E-4 shows the extracted secant value of B for SF 6 (S = 6.) over the applied average axial strain, and compares it to the value obtained from Stanton & Lund (4). The correlation is good and variations lie within -5 %. The reason for the difference can be explained b the simplifing assumption of an isotropic stress state in the theoretical analsis b Stanton & Lund versus a nonlinear plane strain analsis using the FEA. However, the value at zero axial strain, i.e., for the initial stiffness of the unloaded bearing, represents the onl point where linear and nonlinear theories are expected to provide identical results. The difference can be explained from the difference between a real bearing with a ¼ cover laer and the idealized elastomer laer described b Stanton & Lund. Using the length of the shim (8.5 in) instead of the total length of the bearing (9. in) results in a theoretical shape factor of S = 5.67 instead of S = 6.. This has some effect on the inverse calculation using (E-9) with results presented in Figure E-5..6.4. B..8.6 B vs. average axial strain.4 B from FEA. B (Stanton & Lund). -....4.6.8...4 average axial strain (in/in) Figure E-4 B versus average axial strain for bearing with SF 6 - E-9 -
.8.6.4 B...8 B vs. average axial strain.6.4. B from FEA B (Stanton & Lund). -....4.6.8...4 average axial strain (in/in) Figure E-5 B versus average axial strain for bearing with nominal SF 6 and calculation based on theoretical SF 5.67 Figure E-5 indicates that using the length of the shim for the computation of S ields better correlation for the stiffness of the unloaded bearing. However, subsequent nonlinear effects generate underestimation of the bearing stiffness up to 5 % as the axial load increases..8.6.4 B...8 B vs. average axial strain.6.4. B from FEA B (Stanton & Lund). -....4.6.8...4 average axial strain (in/in) Figure E-6 B versus average axial strain for bearing with nominal SF 6 and calculation based on the average SF 5.8 - E- -
Figure E-6 presents a back-calculation of B when using the average length of steel shim and total length of the bearing, leading to S = 5.8. This reduces the error for the loaded bearing without significant error at the unloaded state of the bearing. The above analsis and comparison of Figure E-4 to Figure E-6 demonstrate that using the total length of the bearing when computing the shape factor results in a good balance of the model error over the common loading range of σ = GS. Using the actual length of the shim gives best results for the initial stiffness of the bearing but creates the largest model error as the axial strain, and thus the axial load increases. E... Strip-bearing with SF 9 The bearing analsis is based on an infinite strip bearing as shown in Figure E- and dimensions as given in Table E-. The analsis was performed for axial compression onl. Numerical instabilities limited the analsis to an axial load of P =.kips/in (or an equivalent of P = 464 kips on a 9 in x in bearing).this represents an average stress of σ =.98GS. Table E- Model parameters for SF 9 bearing K 4, psi (,76. MPa) G psi (.69 MPa) E 99.98 psi (.68 MPa) h.laer.5 in n.laer L 9 in Figure E-7 to Figure E-9 show the extracted secant value of B for SF 9 (S = 9.) over the applied average axial strain, and compares it to the value obtained from Stanton & Lund (4) for the same three was of computing the shape factor as in the discussion of Figure E-4 to Figure E-6. The initial zero value for the back-calculated stiffness coefficient is an artifact of the wa data was extracted from the FEA and has no phsical relevance. The model error between linear and nonlinear analsis is still visible, though overall smaller than for SF 6 bearings. Moreover, its variation over the applied axial compression is significantl less that observed for SF 6. The correlation between linear and nonlinear analsis is ver good and variations lie within - %, except when using the length of the steel shim (Figure E-8) when the model error lies within - %. Using the full length of the bearing or the average length of shim and bearings for computing the shape factor result in equall good representation of the axial behavior (Figure E-7 and Figure E-9). - E- -
.4...8 B vs. average axial strain B.6.4. B from FEA B (Stanton & Lund). -......4.5.6.7.8.9. average axial strain (in/in) Figure E-7 B versus average axial strain for bearing with SF 9.6.4.. B vs. average axial strain B.8.6.4. B from FEA B (Stanton & Lund). -......4.5.6.7.8.9. average axial strain (in/in) Figure E-8 B versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 - E- -
.6.4. B..8.6 B vs. average axial strain.4 B from FEA. B (Stanton & Lund). -......4.5.6.7.8.9. average axial strain (in/in) Figure E-9 B versus average axial strain for bearing with nominal SF 9 and calculation based on the average SF 8.75 E... Strip-bearing with SF The bearing analsis is based on an infinite strip bearing as shown in Figure E- and dimensions as given in Table E-4. The analsis was performed for axial compression onl. Numerical instabilities limited the analsis to an axial load of P =.kips/in (or an equivalent of P = 464 kips on a 9 in x in bearing). This represents an average stress of σ =.98GS. Table E-4 Model parameters for SF bearing K 4, psi (,76. MPa) G psi (.69 MPa) E 99.98 psi (.68 MPa) h.laer.75 in n.laer L 9 in Figure E- shows the computed average stress (normalized b GS ) versus the average axial compressive strain. The computed response shows almost perfectl linear response over the entire load histor. Local mesh instabilities near the edge of the steel shims prevent a numerical analsis beond the shown load level. - E- -
avg. stress / GS vs. average axial strain. avg. stress / GS.5..5......4.5.6 average axial strain (in/in) Figure E- σ / GS versus average axial strain for bearing with SF Figure E- shows the obtained value for B as well as the theoretical value after Stanton & Lund (4) over the applied load histor. The equivalent stiffness coefficient obtained from the nonlinear FEA is almost constant, which translates to a virtuall constant secant stiffness and a model error of 7.5 %. This observation further supports the hpothesis of linear behavior of SF bearings at average stress levels of σ GS....8 B vs. average axial strain B.6.4. Bc from FEA Bc (Stanton & Lund). -......4.5.6 average axial strain (in/in) Figure E- B versus average axial strain for bearing with SF Data presented in Figure E- is based on a shape factor calculated from the total length of 9. in which includes both shims and cover laer. Using solel the length of the shims - E-4 -
(i.e., 8.5 in) results in a theoretical shape factor of SF. for the same bearing. The graph in Figure E- shows results of a similar analsis as used for data in Figure E- but based on the theoretical shape factor of SF.. Both figures clearl demonstrate that using the width of the shim and the total width of the bearing, respectivel, for the definition of the shape factor, the linear theor b Stanton & Lund ields lower and upper bounds for the bearing stiffness..4. B..8.6 B vs. average axial strain.4. Bc from FEA Bc (Stanton & Lund). -......4.5.6 average axial strain (in/in) Figure E- B versus average axial strain for bearing with nominal SF and calculation based on theoretical SF.. Using the average length of shim and that of the real bearing for computing the shape factor, results in SF.67 and a stiffness coefficient as shown in Figure E-. Basing the shape factor on the average length of the bearing clearl achieves the best correlation between linear theor and nonlinear FEA at low to moderate axial load. - E-5 -
.4. B..8.6 B vs. average axial strain.4. Bc from FEA Bc (Stanton & Lund). -......4.5.6 average axial strain (in/in) Figure E- B versus average axial strain for bearing with nominal SF and calculation based on the average SF.67. E...4 Summar for the axial stiffness coefficient B The stiffness coefficient B was analzed for three tpes of bearings with SF 6, SF 9, and SF. The listed shape factors are nominal shape factors computed using the total length of the bearing (= 9. in). The analtical equations b Stanton & Lund, however, are based on bearings without cover laer. To assess the significance of the cover laer for the computation of the axial stiffness of a bearing, the identification of B from the numerical analsis was based on three different was of computing the shape factor:. Using the total lengths of the bearing, i.e. including the cover laer (L = 9. in),. Using the actual length of the shim (L = 8.5 in), and. Using the average of shim length and total length of the bearing (L = 8.75 in). This leads to three different coefficients for each bearing if based on the linear theor. For the actual nonlinear analsis, B becomes a function of average axial strain ε or of relative applied load P / AGS = p / LGS = σ / GS. Table E-5 presents a summar of the obtained stiffness coefficients. - E-6 -
Nominal S Table E-5 Comparison of stiffness coefficient l (in) h (in) S lambda B (Stanton & Lund) B for modified definitions of S B (FEA at *GS) Error (%) B (FEA at GS) Error (%) 6 9..75 6..64.8.5..6. 6 8.75.75 5.8.598.8.5..7-6.6 6 8.5.75 5.67.55.8.7-6..49-4. 9 9..5 9..465.. 8..7.6 9 8.75.5 8.75.96.. -..7-4. 9 8.5.5 8.5.8.. -8..9 -.7 9..75..86.4.5 7.9.6 7.7 8.75.75.67.95.5.5 -..5 -. 8.5.75..4.6.5-7.6.6-7.9 The error value gives the relative error of the linear analsis under the assumption that the nonlinear analsis defines the correct answer. Due to the variabilit of the secant stiffness from the nonlinear solution, this error varies. Thus, in Table E-5 the relative error is presented first at zero load, and second at σ =. GS. Best results for each combination are indicated b shaded fields. In most cases, the best correlation between the linear theor b Gent & Meinecke (97), and Stanton & Lund (4) and the present nonlinear analsis is obtained when using the average length of the bearing and the shim for computing the shape factor. Using the linear theor and basing the shape factor on the total length of the bearing results in overestimation of axial stiffness b the linear theor, while using the true length of the shims underestimates the axial stiffness of the bearing. Nonlinear effects become more significant as the shape factor decreases. No significant nonlinear effect is observed over the given range of loads for SF. This statement, however, is onl supported for load levels of σ <. GS on SF bearings, and σ <. GS on SF 6 bearings. Excessive mesh distortion prevents further simulation, but also indicates significant changes in the load bearing mechanism. Experimental evidence clearl demonstrates that the finding of this section do not hold for much higher load levels. - E-7 -
E.. Bending Stiffness Coefficient B r Stanton & Lund (4) showed that the moment-rotation relation for elastomeric bearings can be expressed as M = G I( A r + B r S θ ) t = EI( A r + B r S θ ), (E-) t where M is the total moment on the bearing, EI its linear elastic bending stiffness, A r and B r are stiffness coefficients, S is the shape factor, θ the total rotation on the bearing, and t is the total thickness of all rubber laers (provided the do have identical laer thickness). Solving (E-) for B r ields B r M t = EI θ A r S. (E-) For infinite strip bearings, Stanton & Lund (4) showed that A = 4 /, effectivel reducing (E-) to r B r M t 4 m t = = EWL S θ E L θ 4 S, (E-) with m = M W as moment per unit width, W, of the bearing. / Equation (E-) is used for the identification of B r from nonlinear FE analses of bearings. It could be applied in its total form ielding an equivalent linear secant stiffness, or in an incremental manner ielding a tangent stiffness coefficient. The secant formulation was chosen in order to obtain linear relations for the design procedure. The following four subsections summarize the findings from a series of analses of a SF 9 bearings with rigid end-plates. The aim of this sequence is to answer the question whether or not the rotational stiffness depends on the applied axial load, or if the variabilit (if an were observed) is sufficientl small to justif the use of a linear analsis in a design procedure. This is achieved b studing four distinct load histories for the same bearing. - E-8 -
E... Strip-bearing with SF 9 Rotation at σ =.GS During the first sequence, rotation was applied to the bearing without an initial axial load applied to the bearing. The simulated moments were recorded, and B r was computed using equation (E-). Following identical arguments as presented in section E.., evaluation was performed for SF 9 (= outer length/ t), SF 8.5 (= length of shim/ t), and SF 8.75 (= average length/ t). Figure E-5, Figure E-6, and Figure E-7 show the results from the respective back-calculation versus the value predicted b the linear theor of Stanton & Lund. The nonlinear analsis shows a constant stiffness coefficient for θ / n <.65 rad/laer which indicates perfectl linear behavior. The stiffness appears to change rapidl for θ / n >.65 rad/laer. A simple visual inspection of the computed deformation pattern in that regime reveals a significant hourglass mode pattern over that half of the bearing which is subjected to tension (marked area). Hence, the numericall obtained values are onl valid for θ / n <.65 rad/laer. Figure E-4 shows the deformed FE model at a rotation of θ =.7 rad on the full bearing. The right (tensile) side of the bearing shows the tpical hourglass mode pattern. Figure E-4 Hourglass modes distorting the right side of a bearing with bonded external plates subjected to rotation onl The analsis shows that a shape factor based on the average length of shim and bearing ields the best correlation at virtuall zero model error. Both of the other strategies result in a model error of up to +/-8 %. - E-9 -
. Br vs. Rotation per Laer.5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-5 B r versus average axial strain for bearing with SF 9 at P =. Br vs. Rotation per Laer.5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-6 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =.5 Br vs. Rotation per Laer..5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-7 r B versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P = - E- -
E... Strip-bearing with SF 9 Rotation at σ =.GS During the second sequence, an axial load of P = σ A =. GS A was applied before rotating the bearing. The simulated moments were recorded and B r was computed using equation (E-). Figure E-8, Figure E-9, and Figure E- show the results based on SF 9, SF 8.75, and SF 8.5, respectivel. The nonlinear analsis shows a constant stiffness coefficient for θ / n <.65 rad/laer which indicates perfectl linear behavior. The stiffness appears to change rapidl at θ / n >.65 rad/laer. An analsis of the computed deformation pattern in that regime again reveals significant hourglass mode patterns over that half of the bearing which is subjected to tension. Hence, the numericall obtained values are onl valid for θ / n <.65 rad/laer. The analsis shows that a shape factor based on the total length or on the average length of shim and bearing ields the best correlation at less than % model error. Computing the shape factor based on the length of the shim results in a model error of %. E... Strip-bearing with SF 9 Rotation at σ =.5GS During the third sequence, an axial load of P = σ A =. 5 GS A was applied before rotating the bearing. The simulated moments were recorded and B r was computed using equation (E-). Figure E-, Figure E-, and Figure E- show the results based on SF 9, SF 8.75, and SF 8.5, respectivel. The nonlinear analsis shows a ver slightl increasing stiffness coefficient for the entire range of θ n. No hourglass mode patterns develop under these loading conditions. / The analsis shows that a shape factor based on the total length or on the average length of shim clearl ields the best correlation at -7 % model error. Basing the shape factor on the length of the shim results in a model error of 4 % and thus linear analsis significantl underestimates the rotation stiffness if the wrong shape factor is used in the analsis. - E- -
. Br vs. Rotation per Laer.5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-8 B r versus average axial strain for bearing with SF 9 at P =. GS A..5 Br..5 Br vs. Rotation per Laer..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-9 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =. GS A.5..5 Br..5 Br vs. Rotation per Laer..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E- r B versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =. GS A - E- -
. Br vs. Rotation per Laer.5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E- B r versus average axial strain for bearing with SF 9 at P =.5 GS A.5 Br vs. Rotation per Laer..5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E- B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =.5 GS A.4 Br vs. Rotation per Laer.5..5 Br..5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E- r B versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =.5 GS A - E- -
E...4 Strip-bearing with SF 9 Rotation at σ =.85GS During the forth sequence, an axial load of P = σ A =. 85 GS A was applied before rotating the bearing. This represents the highest load for which a numericall stable solution can be obtained for the presented level of rotation. The simulated moments were recorded and B r was computed using equation (E-). Figure E-4, Figure E-5, and Figure E-6 show the results based on SF 9, SF 8.75, and SF 8.5, respectivel. The nonlinear analsis shows a ver slightl increasing stiffness coefficient for the entire range of θ n. No hourglass mode patterns develop under these loading conditions. / The analsis shows that a shape factor based on the total length or on the average length of shim clearl ields the best correlation at - % model error. Basing the shape factor on the length of the shim results in a model error of 4 % and thus linear analsis significantl underestimates the rotation stiffness if the wrong shape factor is used in the analsis. Neither the numerical results for B r, nor the error analsis shows an significant difference between a load level of P = σ A =. 5 GS A and P = σ A =. 85 GS A. E.. Summar for the rotational stiffness coefficient B r The stiffness coefficient B r was analzed for four different combinations of axial compression and simultaneous rotation. Back-calculation was performed for three different definitions of the shape factor: (a) using the full length of the bearing, (b) using the average between the length of the shim and the full length of the bearing, and (c) using the length of the steel shim to compute S = Aloaded / Asurface = L / t.the analsis shows that () using the full length of the bearing, i.e. including the surface cover for the definition of S ields best results for axiall loaded bearings, () the model error decreases as the axial load increases, and () using the length of the steel shim for the computation of S can cause significant underestimation of the rotational stiffness. - E-4 -
. Br vs. Rotation per Laer.5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-4 B r versus average axial strain for bearing with SF 9 at P =.89 GS A.5 Br vs. Rotation per Laer..5. Br.5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-5 B r versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.75 at P =.89 GS A.4 Br vs. Rotation per Laer.5..5 Br..5..5 Br (from FEA) Br (Stanton & Lund)....4.6.8. Rotation per Laer (rad) Figure E-6 r B versus average axial strain for bearing with nominal SF 9 and calculation based on theoretical SF 8.5 at P =.89 GS A - E-5 -
E.4 Evaluation of Local Shear Strain for Design Purposes Development of a reliable design procedure while maintaining reasonabl low complexit of the required analsis requires. a simple mechanical quantit to correlate experimental data,. a compact and universal relation to compute that measure, and. validation of the procedure against higher order (nonlinear) models. The theories b Gent & Meinecke (97), and Stanton & Lund (4) provide a suitable formulation to satisf requirement, and at the same time suggest γ in the elastomer zx, max at the end of the shims as the answer to requirement. This section will satisf requirement, leaving the answer for the special case of lift-off to section E.5. E.4. Local Effects at the End of the Steel Shims Various levels of mesh refinement were applied to stud local effects near the end of the steel shims. The shims were modeled deformable as mild steel with material parameters as given in subsection E..4. Ver large strains are observed in the elastomer locall near the edge of the shim. The shim itself behaves almost rigid and remains elastic throughout the analsis. Using a linear theor, the sharp edge of the shim will introduce a singularit. Thus, the numerical analsis cannot provide sufficientl high resolution of that local area to provide actual numbers at the ver edge of the shim. The shape of real edges of shims varies between manufacturers from sharp cuts to sanded (slightl rounded). Both the numerical limitations and the variabilit of the real product require some special consideration when analzing and comparing numerical data and results from a linear analsis. This subsection contains a brief discussion of special phenomena observed at the ends of the steel shims. Based on these relations, a strateg for the subsequent analsis is developed. Figure E-7 and Figure E-8 show the distribution of shear strain γ zx = E, with Ε as the in-plane shear component of Green-Lagrange strain. Onl deformation and strain in the elastomer are shown. The undeformed mesh primaril consists of perfectl horizontal and vertical parallel element boundaries. The deformed mesh shows the magnitude of local shear strain even at relativel small axial loads of σ =.5GS (Figure E-7) and σ =.GS (Figure E-8). The figures clearl illustrate that elastomer is squeezed out of the parallel laers and subsequentl is pushed around the edge of the steel shim in a mode of lateral expansion of the elastomer laer. Eventuall, this ver local mode of deformation causes the large strains leading to massive mesh distortion and failure of the numerical analsis. However, the parabolic distribution of horizontal displacements as postulated b the theoretical approach b Gent & Meinecke and also used b Stanton & Lund remains valid at a distance of. in from the edge of the shim. - E-6 -
At σ =.5GS, shear strain reaches extreme values over the last.5 in of the shim (Figure E-7). At σ =.GS, extreme values of shear strain are found concentrated at the edge and to both sides of it (Figure E-8). Figure E-7 Computed shear strain γ = Ε in (in/in) at σ =.5GS at the edge of a steel shim. - E-7 -
Figure E-8 Computed shear strain γ = Ε in (in/in) at σ =.GS at the edge of a steel shim. As the deformation increases, the effect of lateral expansion of the elastomer becomes the dominant mechanism in the vicinit of the end of a shim. This mechanism forces elastomer of the cover laer to expand outward and, due to the nearl incompressible nature of the material, experience lateral contraction of approx. 5 %. Extreme local deformation causes tension both in horizontal and in thickness direction. This creates locall high hdrostatic tension despite an acting compressive stress in vertical direction. Figure E-9 and Figure E- show the magnitude of obtained hdrostatic stress = mean normal stress in tension at σ =.5GS and σ =.GS, respectivel. Areas with compressive mean normal stress are shown in dark gre (largest contiguous area). - E-8 -
Figure E-9 Computed hdrostatic stress σ in (Pa) at σ =.5GS at the edge of a steel shim. The amount of material subjected to hdrostatic tension is larger atσ =.5GS (Figure E-9) and shrinks as loading progresses to σ =.GS (Figure E-). The intensit of the hdrostatic tension, σ, however, increases as the applied load increases. At σ.gs the hdrostatic tension at the end of the shim reaches the magnitude of σ E G. Gent & Lindle (959) showed that internal rupture of elastomers nucleates before this relation is satisfied. This observation has far reaching consequences since it indicates that delamination of the cover laer can be initiated b internal rupture of the elastomer rather than an adhesive bond failure at the interface between elastomer and steel. Visuall, the failure mechanism ma still be observed as delamination since the highest hdrostatic tension is observed in close proximit of the interface. - E-9 -
Figure E- Computed hdrostatic stress σ in (Pa) at σ =.GS at the edge of a steel shim. The intensit of local hdrostatic tension and of local extreme shear strains remains linked to the shear strain in the elastomer at the interface to the steel shim close to the actual edge. Figure E- illustrates the distribution of shear strain in the elastomer along the interface to the steel shim at various load levels. Shear Strain (in/in) 4..... -. -. Shear strain along a shim at different load levels GS =.5 GS =. GS =.5 GS =. GS = GS =~ 4 5 6 7 8 9 -. -4. Position (in) Figure E- Shear strain along the shim at different axial loads for a bearing with SF 9 - E- -