Professional Documents
Culture Documents
Tugas 07 MEH - Chapter 9
Tugas 07 MEH - Chapter 9
METODE ELEMEN
HINGGA
2021
CHAPTER 9
1
Structural Analysis with the Finite Element Method
by Eugenio Onate
CHAPTER 9
MISCELLANEOUS: INCLINED SUPPORTS,
DISPLACEMENT CONSTRAINS, ERROR
ESTIMATION, MESH ADAPTIVITY ETC.
2
9.1 Introduction
1. Boundary Conditions in Inclined Supports
2. Methods to link different element types
3. General constraints in the nodal displacements
Deals with
5. Mesh Symmetries
6. Elastic Supports
Node k
u' ≠ 0
v' = 0
Matrix T transforms
the element stiffness
matrix as and the
equivalent nodal force
vector as
The assembled (symmetric) stiffness matrix can be written before the constraints are enforced as
Note:
X (9.17) is not
symmetric
This equation can be transformed using Eq.(9.14) as
How to fix it:
multiply first the second
row of Eq.(9.16) by and
add the result to the first
X row of Eq.(9.17). Then
multiply the second row
of Eq.(9.17) by −.
Transformation matrix
of Eq. (9.14)
Matrix Stiffness
Solve the problem of Figure 9.6 using the Lagrange multipliers method
The potential energy of the structure Π is now augmented by the penalty function
Obviously, if p = 0 then the constraints (9.24) are satisfied and there is no need to add anything to Π.
Stationarity of Π yields
For αi → ∞ Eq.(9:26) tends to As the α i grow the solution of Eq.(9.26) evolves so that
the constraint equation (9.24) is progressively better satisfied.
The equilibrium equation for finite values of the αi's is deduced from Eqs.(9.22b), (9.24) and (9.26) as
CTαC can be interpreted as the stiffness matrix of a bar element (with α = EA/l) linking nodes 1
and 2. Obviously α → ∞ implies increasing the rigidity of this element and the constraint is
approximated. The new system of equations is
which yields
Clearly the solution “locks” for α → ∞ giving a = 0, unless matrix CTC is singular. This
singularity can be anticipated by observing the number of rows and columns in this matrix. This
problem is similar to the singularity requirement for the shear stiffness matrix in order to
prevent shear locking in Timoshenko beams and Reissner-Mindlin plate and shell elements.
substituting
The smoothing matrix M (e) and the “force" vector g(e) for each element can be assembled in the usual manner to form the
global expression of M and g. Solution of the system (9.55) yields the stresses at all nodes. The stress field has the same
continuity than the interpolation function Ni (typically Co continuity is chosen).
where MD = diag. M. Typically 4 or 5 iterations are needed to obtain a converged solution [ZTZ].
This technique (also called L2 projection of stresses) can be applied to each individual element. A
simpler stress extrapolation procedure can however be implemented at the element level as explained
next.
Fig 9.15
where
The stresses at the end nodes A and B are obtained making s = ± 1/p in (9.59) giving
where
• The main error source in finite element solution is the discretization error
• The error combines unaccuracies introduced by the interpolation and the mesh
chosen
• Enhanced accurate FEM solutions can be achieved by refining the mesh using a
greater number of elements of smaller size
• An alternative approach is keeping the mesh topology constant and using higher
order elements (p method)
• P method are difficult to implement
• The advances in mesh generation procedures has also favored mesh adaptive h-
method
• Square value of the integral error norms can be computed by the sum of element contributions
(9.76)
(9.78)
Percentage of “admissible”
relative error.
Eq (9.77) allows to define a global error parameter as:
(9.79)
(9.80)
(9.82)
(9.84)
Single error parameter for the element: q is the convergence rate of local
error parameter depending on the
(9.83)
mesh optimality criterion chosen.
(9.79)
(9.83) Comparing
(9.89)
(9.82)
85