Charilaos Mylonas - Personal Introduction

Charilaos Mylonas - Personal Introduction
PhD Candidate - Charilaos Mylonas
Institute of Structural Engineering Method of Finite Elements I 1

Personal Introduction
Studies
Civil Engineering - Aristotle University of Thessaloniki

Computational Science and Engineering - ETH Zurich
Research interests/research focus
Computational techniques for uncertainty quantification

Numerical solution of PDEs (Including Finite Elements)
Current research focus: uncertainty propagation for Wind Turbine
composite blade fatigue analysis

Previous FE-related projects - motivation
Homogenization for Composites
Composite elastic response cannot be treated in the

homogeneous elasticity framework (”Lamé” parameters, λ, µ).
Anisotropic elastic properties can be approximated from the
solution of elasticity-like problems in the microscale.
PDE Boundary Value Problem using elastic constitutive
relations and assuming,
(0) (1)
ui = ui + ui + 2 u(2) + · · ·
where → 0 1 . Displacement ends up being decomposed in a
fast spatially varying periodic component u(1) that we can
analyse in the microscale, induced by the periodicity of the
microstructure, and a ”slow” component that we can analyse in
the macroscopic scale.
1
This is a technique called perturbation method. Introduced for the analysis
of periodic structures (composites) in [BLP78]
Homogenization for Composites
Weak form (aka - Variational form) 2 :
∂χmn ∂νi
Z Z
∂Dijmn
Dijkl k dy = νj dy
Ω ∂yl ∂yj Ω ∂yi
where Ω our microscopic periodically repeating cell, Dijkl
constitutive stiffness tensor3 and χ a function that that satisfies
(0)
(1) ∂uk (1)
ui = χkl
i + ũi (x)
∂xl
where x = {x1 , x2 , x3 } the coordinates in the macro-scale.
2
An expression between integrals that when it is satisfied, the PDE boundary
value problem is satisfied. In practice it’s minimized (we search for a set of ai
that make this expression stationary) not ”satisfied”.
3
For homogeneous materials can be represented by Lamé parameters
Homogenization for Composites - Ritz-Galerkin method
We want to approximate χmn

i .
We take νi to be a linear combination of functions
PN (i) (i)
(Ritz/Galerkin idea ) νi (y) = m=1 am Ψm (y) and the same
for χmn
i .
By passing these approximations in the variational form, a linear
system that we can solve with standard linear algebra arises
however, as you will learn, by a special choice of the Ψ
functions the linear system has some nice properties.

Homogenization for Composites - Finite Element solution
This system is sparse (most of the elements are zero) when

using functions that overlap only at a small region. Idea
attributed to Courant - That’s what we call today Finite
Elements.
Solution with FE:
The solution for one component of χ consists of approx. 70000

local and piecewise linear functions (aka - hat functions).
Try inv(rand(70000,70000)) on Matlab... Due to Courant and
sparse linear algebra nobody ever needs to do that for solving
PDEs.
Homogenization for Composites - Finite Element solution
Why homogenization? Great computational savings, while being

able to retrieve information about the micro-stresses in the
composite.
Nothing more than the mathematically consistent way to
compute macroscopic properties for periodic structures.
For more details on the computational treatment of the
homogenization problem refer to [CTN01].
Many big names introduced here (Ritz, Galerkin, Courant). For
a historical review, [GW12] is highly recommended.
For students interested in computational methods in general 4
[Str08]
4
not strictly oriented to this course but for general education on numerical
methods in engineering
Hyperelastic Materials
Architects consider soft, air-pressure actuated structures for

facades
In order to quickly test new designs, and assess their mechanical
behavior, FE models were developed:

Current project: Fatigue Assessment of Wind Turbine
Blades
Composite materials - thin cross-sections
Special purpose tools for the estimation of warping and shear

flows are used. 5
Time-dependent FE analysis for aerodynamic loads on beams
Cross-section analysis on beam resultants:
5
Recall theory of thin-walled beam analysis. Here slightly different due to
composite anisotropy. You may refer to [GBM+ 83] and [YHH12] for more
information on the treatment.
Current project: Fatigue Assessment of Wind Turbine
Blades
Issues with simulation:
A problem where ”less is more”

a complete 3D FE treatment practically unfeasible (high
computational times - time dependent simulations do not
scale 6 )
So-called mixed 7 FE-formulations have found application for
increased accuracy in beam-type modelling.
6
Solving with 10 CPUs does not make (in general) the solution 10 times
faster. Cannot straightforwardly benefit from advances in hardware.
7
variables of different ”nature” enter the weak formulation - such as velocities
and rotations, instead of displacements
A step back
For analysis of wind turbine blades back to Ritz/Galerkin:

Displacements on beams, are generally smooth
a piecewise linear approximation may not be the best approach
for a basis function set...
In the context of the ”Geometrically Exact Beam Theory” other
choices for the basis functions may be investigated.
For the intrigued students, please do refer to [PA11].
In Wind-turbine simulation codes Legendre polynomials already
considered.

Other interests
Uncertainty Quantification:
Probabilistic treatment8 of uncertainty with non-intrusive 9
methods
Optimization:
Partial Differential Equation constrained shape optimization
Genetic algorithms and Structural optimization
8
Meaning doing something more efficient than running many times with
different parameters (Monte-Carlo)
9
meaning not changing existing FE code
Thank you for your attention!
Thank you for your attention! - we wish you a productive and

creative semester!

References I
Alain Bensoussan, Jacques-Louis Lions, and George

Papanicolaou.
Asymptotic analysis for periodic structures, volume 5.
North-Holland Publishing Company Amsterdam, 1978.
Peter W Chung, Kumar K Tamma, and Raju R Namburu.
Asymptotic expansion homogenization for heterogeneous media:
computational issues and applications.
Composites Part A: Applied Science and Manufacturing,
32(9):1291–1301, 2001.
Vittorio Giavotto, Marco Borri, Paolo Mantegazza,
G Ghiringhelli, V Carmaschi, GC Maffioli, and F Mussi.
Anisotropic beam theory and applications.
Computers & Structures, 16(1-4):403–413, 1983.

References II
Martin J Gander and Gerhard Wanner.

From euler, ritz, and galerkin to modern computing.
Siam Review, 54(4):627–666, 2012.
Mayuresh J Patil and Matthias Althoff.
Energy-consistent, galerkin approach for the nonlinear dynamics
of beams using intrinsic equations.
Journal of Vibration and Control, 17(11):1748–1758, 2011.
Gilbert Strang.
Mit ocw 18.085- computational science and engineering i.
MIT OpenCourseWare: Massachusetts Institute of Technology,
2008.

References III
Wenbin Yu, Dewey H Hodges, and Jimmy C Ho.

Variational asymptotic beam sectional analysis–an updated
version.
International Journal of Engineering Science, 59:40–64, 2012.

Charilaos Mylonas - Personal Introduction

Uploaded by

Copyright:

Available Formats

Charilaos Mylonas - Personal Introduction

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Charilaos Mylonas - Personal Introduction

Uploaded by

Copyright:

Available Formats

Charilaos Mylonas - Personal Introduction

PhD Candidate - Charilaos Mylonas

Institute of Structural Engineering Method of Finite Elements I 1

Civil Engineering - Aristotle University of Thessaloniki

Research interests/research focus

Computational techniques for uncertainty quantification

Institute of Structural Engineering Method of Finite Elements I 2

Composite elastic response cannot be treated in the

Weak form (aka - Variational form) 2 :

Homogenization for Composites - Ritz-Galerkin method

We want to approximate χmn

Institute of Structural Engineering Method of Finite Elements I 5

This system is sparse (most of the elements are zero) when

The solution for one component of χ consists of approx. 70000

Homogenization for Composites - Finite Element solution

Why homogenization? Great computational savings, while being

Architects consider soft, air-pressure actuated structures for

Institute of Structural Engineering Method of Finite Elements I 8

Special purpose tools for the estimation of warping and shear

Issues with simulation:

A problem where ”less is more”

For analysis of wind turbine blades back to Ritz/Galerkin:

Institute of Structural Engineering Method of Finite Elements I 11

Thank you for your attention! - we wish you a productive and

Institute of Structural Engineering Method of Finite Elements I 13

Alain Bensoussan, Jacques-Louis Lions, and George

Institute of Structural Engineering Method of Finite Elements I 14

Martin J Gander and Gerhard Wanner.

Institute of Structural Engineering Method of Finite Elements I 15

Wenbin Yu, Dewey H Hodges, and Jimmy C Ho.

Institute of Structural Engineering Method of Finite Elements I 16

You might also like