Hi all
In nonlinear finite element analysis references,such as Crisfield, Arc-Length is one of solution algorithm for nonlinear equations, but in Opensees Arc-Length is one of integrator methods. Is there any difference between them? I want to use Arc-Length method for algorithm command, and Newmark method for integrator command. Is it possible?
Thanks.
Arc-Length
Moderators: silvia, selimgunay, Moderators
Arc-Length
Last edited by imancivil on Mon Dec 01, 2008 8:40 am, edited 1 time in total.
Iman Mansouri
PhD Candidate
University of Kerman
PhD Candidate
University of Kerman
it is not possible. Arc-Length is a continuation method attempting to solve (9.15 in Crisfield)
g(P,lambda) = Qi(P) - lambda *Qref, Qi the internal forces a function of fixed external forces P and displacements, and Qref the reference loads. arclength it is attempting to determine the lambda. the loadfactor to be applied to external reference loads to solve this
condition and the other conditions of the continuation method, in arc-length case this condition is based on the arc-length.
In OpenSees we call these static integrators, and we use solution algorithms such as Newmark to iterate to the solution.
g(P,lambda) = Qi(P) - lambda *Qref, Qi the internal forces a function of fixed external forces P and displacements, and Qref the reference loads. arclength it is attempting to determine the lambda. the loadfactor to be applied to external reference loads to solve this
condition and the other conditions of the continuation method, in arc-length case this condition is based on the arc-length.
In OpenSees we call these static integrators, and we use solution algorithms such as Newmark to iterate to the solution.
A follow-up question on this: Newmark method is a direct integration method which finally gives a nonlinear equations with effective stiffness and load vector, so in order to solve it, an advancing scheme still need to be specified, say 'arc-length',' loadcontrol' or 'displacement control'. But opensees did not ask for it when performing the transient analysis, so is there a default setting regarding this issue?fmk wrote:it is not possible. Arc-Length is a continuation method attempting to solve (9.15 in Crisfield)
g(P,lambda) = Qi(P) - lambda *Qref, Qi the internal forces a function of fixed external forces P and displacements, and Qref the reference loads. arclength it is attempting to determine the lambda. the loadfactor to be applied to external reference loads to solve this
condition and the other conditions of the continuation method, in arc-length case this condition is based on the arc-length.
In OpenSees we call these static integrators, and we use solution algorithms such as Newmark to iterate to the solution.
ArcLength, DisplacementConrol and others are Static integrators. Newmark, HHT are dynamic integrators. The integrators just define what the equilibrium equation is at each point in time.
In OpenSees you specify in addition o others the Integrator and the Algorithm. It is the solution algorithm that handles solvuing the nonlinear equation. Example algorithms are Linear, NewtonRaphson, ModifiedNewton, etc.
In OpenSees you specify in addition o others the Integrator and the Algorithm. It is the solution algorithm that handles solvuing the nonlinear equation. Example algorithms are Linear, NewtonRaphson, ModifiedNewton, etc.
Thanks for the reply, Frank. My understanding was Newton-family algorithms provide a corrector phase when solving the nonlinear equation, however, certain advacing scheme (let the solution moves forward) has to be provided too. Maybe Opensees used a different way of handling this issue.fmk wrote:ArcLength, DisplacementConrol and others are Static integrators. Newmark, HHT are dynamic integrators. The integrators just define what the equilibrium equation is at each point in time.
In OpenSees you specify in addition o others the Integrator and the Algorithm. It is the solution algorithm that handles solvuing the nonlinear equation. Example algorithms are Linear, NewtonRaphson, ModifiedNewton, etc.