Energy Approach to Convergence Acceleration of Step-
By-Step Iterative Methods in Finite Element Analysis of
Geometrically Nonlinear Structures Volume 1 - Issue 5
Vladimir P Agapov*
Department of Civil Engineering, National Research Moscow State University, Russia
Received: December 14, 2021 Published: January 25, 2022
Corresponding author: Department of Civil Engineering, National Research Moscow State University, Russia
One of the problems arising in the analysis of nonlinearly
deformable structures by
the finite element method is the acceleration of the convergence
of step-iterative procedures. In this paper, a method for convergence
accelerating is considered, based on the use of both static and
energy correction of the solution.
Equilibrium equations of a structure for a load step can be
written in the form [1]:
where
- the stiffness matrices of the zero, first
and second orders
- the matrices of initial displacements and stresses {Δu}- the vector of increments of nodal displacements
{ΔP}- the vector of increments of nodal forces, respectively.
Matrices for an individual finite element are written as follows
[2]:
Where
are the direct and differential stiffness
matrices of the i-th order, n is the number of degrees of freedom of
a finite element.
Equation (2) means that the j-th row of the matrix can be found
as the product of the transposed vector of nodal displacements
by the first derivative of NLi
with respect to the j-th degree of freedom.
When solving a nonlinear static problem in increments and
using modified Lagrangian coordinates, equation (1) is written in
the form:
We will solve equation (3) by an iterative method of additional
loading, which is equivalent to using the modified Newton-Raphson
method:
where j - the number of the loading step, i - the iteration number
at this step.
To accelerate the convergence of the iterative process, we use
the energy relations. The displacement vector determined from the
equilibrium equations (4) must also satisfy the energy conservation
law. For each loading step, you can write
where
the work of external and internal forces
of the initial state on initial displacements,
- the work
of external and internal forces of the initial state on additional
displacements
- the work of additional external and
internal forces on additional displacements.
For the initial state, the energy conservation law is observed.
Hence
Since the initial state is in equilibrium, then, considering the
additional displacements to be sufficiently small, based on the
principle of possible displacements, we can write
Considering (5) - (7), we obtain:
We require that the solution obtained from (4) at the i-th
iteration satisfies relation (7), i.e.
To achieve this, we introduce the correction factor “c” as follows:
Let’s calculate the work of internal forces. As shown in [1]:
Where
The work of external forces can be found like this:
Using relations (10) - (13), we arrive at the algebraic equation:
whose coefficients are equal:
Equation (14), as a rule, has one positive root. However, if there
are several such roots, then you need to choose the closest to one
in order to provide the smallest number of equilibrium iterations.
The described above method of convergence acceleration was
realized in computer program PRINS [3]. The effectiveness of the
method has been proven in practice.
Agapov VP (1984) Osnovnye sootnosheniya MKE v staticheskih i dinamicheskih raschetah geometricheski nelinejnyh konstrukcij // Stroitel'naya mekhanika i raschet sooruzhenij. - N 5 C pp. 43-47.
Agapov VP Uchet geometricheskoj nelinejnosti v staticheskih i dinamicheskih raschetah konstrukcij metodom konechnyh elementov// Uchenye zapiski CAGI (5): 90-98.
Agapov VP (2005) Metod konechnyh ehlementov v statike, dinamike i ustojchivosti konstrukcij, Izd-vo ASV, Moscow, Russia.