Scribd
Upload
123 views

Adaptive Refinement in Vibrational Analysis and Isogemetric Analysis

This document discusses principles of adaptive mesh refinement in vibrational analysis and new developments. It covers topics like error indicators, parallelization, coupling of dynamics and physics, handling time-stepping, and classifications of adaptive methods like h-refinement and p-refinement. It also summarizes concepts like isogeometric analysis, non-uniform rational B-splines (NURBS), and applications of adaptive meshing like modeling crack propagation.

Uploaded by

Abhishek Kumar
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
123 views

Adaptive Refinement in Vibrational Analysis and Isogemetric Analysis

This document discusses principles of adaptive mesh refinement in vibrational analysis and new developments. It covers topics like error indicators, parallelization, coupling of dynamics and physics, handling time-stepping, and classifications of adaptive methods like h-refinement and p-refinement. It also summarizes concepts like isogeometric analysis, non-uniform rational B-splines (NURBS), and applications of adaptive meshing like modeling crack propagation.

Uploaded by

Abhishek Kumar
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 22

Adaptive Refinement in

Vibrational Analysis & New


Developments
ABHISHEK KUMAR
QHS 019
Principles of mesh refinement
Run the same model on grids with different space/time resolutions
Required for the embedding:
A time integration algorithm
Grids interactions
Required for the adaptivity:
A refinement criterion
An efficient grids initialization
procedure

Error Indicators: This is the mechanism by which an adaptive
mesh generator decides to coarsen or refine the mesh. The question we
need to know a priori is what sorts of processes do we wish to track (error
indicators try to equi-distribute the error in a global sense). The error
indicator can do its job if the spatial and temporary resolutions have high
fidelity. This means that spatial, temporal, and boundary conditions should
have a balanced order of accuracy.
Parallelization/Domain Decomposition: Modifying the
data structures dynamically slows the computations. E.g., the domain
decomposition needs to be a direct by-product of the adaptive mesh
generator. A good first candidate for AM is statically adaptive grids where
the grid is modified and held fixed for the entire simulation. This must work
well before moving onto dynamically adaptive grids.



Some Standing Issues for Adaptive
Methods

Coupling of Dynamics and Physics: Sub-grid scale parameterization is
notoriously inconsistent meaning that changing the grid resolution
changes the results. Also, the dynamics must use the proper
approximations for the smallest scales.. This has direct effects on the
time-integration strategies.

Handling Time-stepping: By definition, multi-scale problems have
processes occurring at all spatial scales that then require special
treatment with respect to time in order to run efficient simulations (stiff and
non-stiff parts).

For explicit methods, the time-step has to be small enough to maintain stability.
For semi-implicit methods, one must still must adhere to the time-step stability
of the nonlinear terms.
High-accuracy yet efficient time-integrators can be obtained. These methods
are easily modified to yield variable order in time with adaptive time-stepping.
Some Standing Issues for Adaptive Methods
(Continued)
R-refinement: The number of points remains constant for all time; however, the
position of these points move.

H-refinement: The number of grid points changes with time.

P-refinement: The number of grid points and their locations remain the same, the
number of modes used to construct the solution changes. This is only possible with
Galerkin methods.

HP-refinement: Combining AMR (h-refinement) and (p-refinement)
(e.g., Bernard, Chevaugeon, Legat, Deleersnijder, Remacle 2007).

K-refinement

Key Point: All these methods move points closer together so that special handling
of the time-step must be considered.
General Classification of Adaptive Methods
6
Principles of mesh refinement
G
0
G
1
G
2
interpolation
1
6
5 4 3
2
11
10
9 8 7 13 12
update
Time integration algorithm

Mode Dependent Error Estimator for
Vibration Analysis
8
Spatial Error Estimator Mode Superposition Technique
Generalized coordinate transformation,
where o is an n x m eigenvector (mode shape) matrix

The time history of dynamic stresses
where o(t) = the stress components for K
th
mode

The spatial error estimator at each time step in energy norm


where o
*
(t) = improved/smoothed stress and
N = total number of elements

The error indicator at each time step

where is the exact energy norm at time t(sec)

Compute the element error indicators at the time step at which
is maximum
) t ( q q ) t ( x
k
m
1 k
k

o = o =
=


o = = o
=
) t ( q B D x(t) B D ) t (
k
m
1 k
k
( ) ( )
2 1
N
A
* 1 - *
dA ) t ( ) t ( D (t) - ) t ( ) t ( e


o o o o =
100 x
u(t)
e(t)
(%) ) ( = t n
) t ( u
) t ( n
Explicit Time-Integrators

Semi-Implicit (IMEX) Time-Integrators

Lagrangian Time-Integrators

Fully-Implicit Time-Integrators
Time-Stepping Strategies
Implementing Boundary Conditions

For explicit time-integrators,
implementing BCs is trivial (a
posteriori).

For semi-implicit/implicit time-
integrators, the BCs have to be
incorporated into the time-integrator.

BCs can be generalized quite
naturally via Lagrange multipliers as
such:
Where tau satisfies all BCs
including NFBC and NRBCs
We are currently extending high-
order BCs into this formulation.

cq
ct
= S(q) +tn
cq
ct
= S(q)
q
n+1
= P
(BC)
q
Isogeometric Analysis
of Structural
Vibrations
TJR Hughes
JA Cottrell





Prerequisites: NURBS, Isogeometric Analysis, Structural Vibration, Invariant frequency
spectra,
Independent evolution of CAD and FEM
CAD (NURBS) and Finite Elements evolved in different communities
before electronic data exchange
FEM developed to improve analysis in Engineering
CAD developed to improve the design process
Information exchange was drawing based, consequently the
mathematical representation used posed no problems
Manual modelling of the element grid
Implementations used approaches that best exploited the limited
computational resources and memory available.
FEA was developed before the NURBS theory
FEA evolution started in the 1940s and was given a rigorous
mathematical foundation around 1970 (e.g, ,1973: Strang and Fix's An
Analysis of The Finite Element Method)
B-splines: 1972: DeBoor-Cox Calculation, 1980: Oslo Algorithm

Why have NURBS not been used in FEA?
FEA was developed before the NURBS theory
NURBS and Finite Elements evolved in different communities before
electronic data exchange
It was agreed that higher order representations in most cases
did not contribute to better solutions
Current computers have extreme performance compared to earlier
computers. Allows more generic solutions.
Mathematical representation in CAD and FEA chosen based on
what was computationally feasible.
We needed someone with high standing in FEA to promote the idea
of splines in analysis
Tom Hughes did this in 2005
The Computer Aided Design Community has adopted the idea
A new drive in spline research after 10 quite years
14
The Isogeometric simulation process
Create geometry
CAD model
Translate model
(can involve
simplification
and
approximaton)
Isogeometric
model
Add boundary
conditions,
properties
Refinement
(the model is
essencially the
same, only
enriched with
more information)
Isogeometric
mesh
Perform
analysis
Visualization
Result
Update geometry
(which version of the
geometry should be
updated?)
An analysis suitable model
USER
Non-uniform rational B-spline
Non-uniform rational basis spline (NURBS) is a
mathematical model commonly used in computer
graphics for generating and representing curves
and surfaces which offers great flexibility and
precision for handling both analytic (surfaces
defined by common mathematical formulae) and
modeled shapes.
Development of NURBS made possible a
mathematically precise representation of freeform
surfaces like those used for ship hulls, aerospace
exterior surfaces, and car bodies, which could be
exactly reproduced whenever technically needed.
NURBS are commonly used in CAD, CAM, and
CAE. NURBS tools are also found in various 3D
modeling and animation software packages.
Why are splines important to Isogeometric
analysis?
Splines are polynomial, same as Finite Elements
B-Splines are very stable numerically
B-splines represent regular piecewise polynomial structure in a
more compact way than Finite Elements
NonUniform rational B-splines can represent elementary curves and
surfaces exactly. (Circle, ellipse, cylinder, cone)
Efficient and stable methods exist for refining the piecewise
polynomials represented by splines
Knot insertion (Oslo Algorithm, 1980, Cohen, Lyche, Riesenfeld)
B-spline has a rich set of refinement methods

Adaptive Mesh
Applications
Adaptive Applications
Highly adaptive and irregular applications
Amount of work per task can vary
drastically throughout the execution
Computations in the interesting regions
of the domain larger than for other
regions
It is difficult to predict which regions will
become interesting
AMR Applications
An example of such applications is Parallel
Adaptive Mesh Refinement (AMR) for multi-
scale applications
Adaptive Mesh Mesh or grid size is not
fixed as in Laplace/Jacobi, but interesting
regions are refined to form finer level
grids/mesh
E.g.: to study crack growth through a
macroscopic structure under stress

AMR Applications Crack
propagation
Such a system is subject to the laws of
plasticity and elasticity and can be
solved using finite element method
Crack growth forces the geometry of
the domain to change
This in turn necessitates localized
remeshing.

AMR Applications- Adaptivity
Adaptivity arises when advances crosses
from one subdomain to another
It is unknown in advance when or where
the crack growth will take place and which
subdomains will be affected
The computational complexity of a
subdomain can increase dramatically due to
greater levels of mesh refinement
Difficult to predict future workloads

You might also like