Force-based Element vs.

Displacement-based Element
Vesna Terzic UC Berkeley

December 2011

Agenda
•  •  •  •  •  •  Introduction Theory of force-based element (FBE) Theory of displacement-based element (DBE) Examples Summary and conclusions Q & A with web participants

which is especially important in the presence of distributed element loads (girders with high gravity loads).Introduction •  Contrary to concentrated plasticity models (elastic element with rotational springs at element ends) force-based element (FBE) and displacementbased element (DBE) permit spread of plasticity along the element (distributed plasticity models). . •  Distributed plasticity models allow yielding to occur at any location along the element.

.Introduction •  OpenSees commands for defining FBE and DBE have the same arguments: element forceBeamColumn $eleTag $iNode $jNode $numIntgrPts $secTag $transfTag element displacementBeamColumn $eleTag $iNode $jNode $numIntgrPts $secTag $transfTag •  However a beam-column element needs to be modeled differently using these two elements to achieve a comparable level of accuracy. •  In order to enhance your understanding of these two elements and to assure their correct application I will present the theory of these two elements and demonstrate their application on two examples. •  The intent of this seminar is to show users how to properly model frame elements with both FBE and DBE.

v 3 v = af u T p = a GeometricTran f q v q2.v 1 p q3.v 2 q Basic System .Transformation of global to basic system up Element u q1.

constant axial deformation and linear curvature distribution are enforced along the element length (exact only for prismatic linear elastic elements) •  Mesh refinement of the element is needed to represent higher order distributions of deformations. •  To approximate nonlinear element response.Displacement-based element •  The displacement-based approach follows standard finite element procedures where we interpolate section deformations from an approximate displacement field then use the PVD to form the element equilibrium relationship. exact curvature .

Formulation of DBE (1) For 2D element: Displacement interpolation: Assuming constant axial deformation and linear curvature distribution along the element length we get: N(x) – matrix of shape functions (Hermitian polynomials) Strain-displacement relationship: .

Formulation of DBE (2) Element basic forces: (PVD is used to formulate equilibrium between s(x) and q) This is weak equilibrium that leads to error in force boundary conditions. internal forces are not in equilibrium with element basic forces. To assemble tangent stiffness matrix of the system we need to know tangent stiffness matrix of each element: .sectional stiffness . Thus.material tangent stiffness .

v 2 q Basic System e s Section ! ! σ Material ε .v 1 v q2.Formulation of DBE (3) q3.v 3 q1.

Equilibrium between element and section forces is exact. •  Section forces are determined from the basic forces by interpolation within the basic system.Force-based element •  The force-based approach relies on the availability of an exact equilibrium solution within the basic system of a beam-column element. -  Interpolation comes from static equilibrium and provides constant axial force and linear distribution of bending moment in the absence of distributed element loads. which holds in the range of constitutive nonlinearity. •  PVF is used to formulate compatibility between section and element deformations: .

Formulation of FBE Force interpolation: From the equilibrium in the undeformed configuration of a free body that comprises the portion of the element between node i and the section at x we get in the absence of element loading: b(x) – matrix of force interpolation functions Compatibility between section and element deformation: (compatibility is satisfied in integral form at the element ends rather than for all values of x and derived using PVF) .

Formulation of FBE (2) To assemble tangent stiffness matrix of the system we need to know tangent stiffness matrix of each element: Tangent flexibility matrix of element: .sectional flexibility .

v 3 q1.Formulation of FBE (3) q3.v 1 v q2.v 2 q r Basic System Applied and resisting sectional forces need to be in equilibrium – Newton iterative procedure External and internal element deformations need to be in equilibrium – solved by Newton iterative procedure e s For current q: (we start with the initial guess q0 and update with ) Section r For current e: (we start with the initial guess e0 and update with ) ! ! σ Material ε .

Filippou.Example 1 – Steel Beam Neuenhofer. C. and F. . 123(7): 958-966. Journal of Structural Engineering. A. Evaluation of Nonlinear Frame Finite Element Models. (1997)..

Example 1 – Results Rotation error (node B) Global response Curvature error (node A) Local response FB BCE FB BCE DB BCE DB BCE .

does not involve discretization error but only the numerical error.) An error less than 2% is obtained for both global and local response quantities with only one element and 7 IPs. . (This is due to the fact that FBE uses the exact force interpolation functions and thus.Example 1 – FBE Rotation error (node B) Global response Curvature error (node A) Local response FBE FBE •  •  Accuracy of the solution can be improved by either increasing the NIP (preferable from a computational standpoint) or the number of elements.

involve both discretization and numerical error. 8 elements are required to reduce the rotation error to ~0. .Example 1 – DBE Rotation error (node B) Global response Curvature error (node A) Local response DBE DBE •  •  Accuracy of the solution can only be improved by increasing the number of elements (not by increasing the NIPs). and 16 elements are required to reduce the curvature error to 3%. This is due to the fact that DBE uses displacement interpolation functions that approximate the exact solution and thus.

. both local and global quantities converge fast with increasing NIPs.Example 1 .Summary •  Accuracy of the solution can be improved: –  for FBE.. rotations). •  In case of DBE. higher derivatives converge slower to the exact solution and thus.g. only by increasing the number of elements. •  In case of FBE. –  for DBE.. by either increasing the NIPs (preferable from a computational standpoint) or the number of elements.g. curvature) requires a finer finiteelement mesh than the accurate determination of global response quantities (e. accurate determination of local response quantities (e.

PEER 1998/01 (Column 415)) 8 -0 .Example 2 – Bridge Column •  Bridge column (Lehman & Moehle.

Example 2 .Loading protocol 8 6 4 2 Ductility 0 -2 -4 -6 -8 0 1000 2000 3000 4000 5000 .

To provide better accuracy of local strains NIPs is chosen such that integration weight of the end node matches the plastic hinge length.FB 200 Force [kips] Force [kips] Force [kN] 0 -20 -40 -60 -80 -8 0 -100 -200 -300 -6 -4 -2 0 2 4 6 8 0 -100 -200 Displacement [in] Displacement [in] Force [kN] 20 100 .FB. 1 Ele 15 300 Experimental Analytical .Example 2 – Model calibration •  The column model is calibrated using force-based element with 5 integration points. Displacement [cm] -20 80 60 40 -15 -10 -5 0 5 10 15 20 300 200 100 Displacement [cm] -15 60 40 20 0 -20 -40 -60 -6 -4 -2 0 2 4 -300 6 -10 -5 0 5 10 Experimental Analytical .

FBE vs. 1 Ele 200 60 40 100 Force [kips] Force [kips] Force [kN] 0 0 -20 -40 -60 -80 0 -100 -200 -300 -400 -100 -200 -100 -6 -4 -2 0 2 4 6 Displacement [in] Displacement [in] Force [kN] 20 100 .Example 2 – FBE vs.FB. DBE •  The response will change significantly by replacing the forcebased beam-column element with the displacement-based beamcolumn element. Experiment (NIP = 3) Displacement [cm] -5 0 5 10 15 400 300 200 Experimental Analytical .DB. 1 Ele 15 300 -15 100 80 -10 DBE vs. Experiment (NIP = 5) Displacement [cm] -15 60 40 20 0 -20 -40 -60 -6 -4 -2 0 2 4 -300 6 -10 -5 0 5 10 Experimental Analytical .

4 Ele 200 40 20 0 -20 -40 -60 200 Force [kips] Force [kN] Force [kN] 0 -20 -40 -60 -80 0 -100 -200 0 -20 -40 0 0 -100 -100 -300 -400 -200 -200 -60 -6 -4 -2 0 2 4 -300 6 -100 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 -300 6 Displacement [in] Displacement [in] Displacement [in] Force [kN] 20 Force [kips] Force [kips] 100 20 100 100 .DB. Experiment Column modeled with 4 elements Displacement [cm] -15 60 -10 -5 0 5 10 15 300 -10 -5 0 5 10 Experimental Analytical .DB.Example 2 . Experiment Column modeled with 1 element Displacement [cm] -15 100 80 60 40 -10 -5 0 5 10 15 400 300 DBE vs. 2 Ele 15 300 Experimental Analytical . DBE vs.DB. 1 Ele Experimental Analytical .DBE •  With the increase of number of DBE the analytical prediction better matches the measured response of the column. Experiment Column modeled with 2 elements Displacement [cm] -15 60 40 200 DBE vs.

Local response quantities could not be compared due to the lack of experimental data. However. Experiment Column modeled with 1 element Displacement [cm] -15 60 40 20 0 -20 -40 -60 -6 -4 -2 0 2 4 -300 6 -10 -5 0 5 10 Experimental Analytical . FBE vs. Experiment Column modeled with 4 elements Displacement [cm] -15 60 200 -10 -5 0 5 10 Experimental Analytical . it is advisable to use more then 4 DBE when predicting local response quantities. 4 Ele 15 300 40 20 0 -20 -40 -60 -6 -4 -2 0 2 4 200 100 Force [kips] 100 Force [kips] Force [kN] 0 0 -100 -100 -200 -200 -300 6 Displacement [in] Displacement [in] Force [kN] .DB.FB.Example 2 . 1 Ele 15 300 DBE vs. the column had to be modeled with either 1 FBE or 4 DBE.Summary •  •  To match the measured column response.

FBE generally improves global and local response without mesh refinement. by either increasing the NIPs (preferable from a computational standpoint) or the number of elements. both local and global quantities converge fast with increasing NIPs. In case of DBE. –  for DBE. only by increasing the number of elements.Summary and conclusions •  •  FBE and DBE can not be modeled in the same way as they inherently differ from each other Accuracy of the solution can be improved: –  for FBE. higher derivatives converge slower to the exact solution and thus. accurate determination of local response quantities requires a finer finiteelement mesh than the accurate determination of global response quantities. Although computationally more expensive. •  •  •  •  . In case of FBE. NIPs of a FBE has to be chosen such that integration weights at locations of plastic hinges match the plastic hinge lengths. To accurately capture local response of elements whose plastic hinges locations and lengths can be estimated.

Sign up to vote on this title
UsefulNot useful