Вы находитесь на странице: 1из 77

Introduction.

Industrial applications
Electromagnetic modelling
Magnetic field formulation
Magnetic field/magnetic potential formulation
Numerical solution

Numerical methods in electromagnetism and applications

Alfredo Bermúdez de Castro


Departamento de Matemática Aplicada, Universidade de Santiago de Compostela. Spain

Colloquium del Departamento de Matemáticas de la


Universidad Carlos III
Madrid April 17, 2012

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Magnetic field formulation
Magnetic field/magnetic potential formulation
Numerical solution

Outline
Introduction. Industrial applications
Electrodes for electric arc furnaces
Induction furnaces
Electric motors
Microwave heating
Electromagnetic modelling
Maxwell’s equations
Constitutive laws
Harmonic eddy currents model
Magnetic field formulation
Strong problem
Weak formulation
Magnetic field/magnetic potential formulation
Scalar magnetic potential
Weak formulation
Numerical solution
Finite element approximation
Discretized problems. Error analysis
Computer implementation
Numerical experiments

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Industrial applications of electromagnetism

Electrodes for electric arc furnaces


Induction furnaces
Electric motors
Microwave heating

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electrodes for electric arc furnaces

Contract with Ferroatlántica I+D company. Also financed by


the the Spanish government), FEDER and Xunta de Galicia
FA is interested in silicon production
FA invented a new compound electrode called ELSA in the
1990. It is the world leader in the sector of silicon of
metallurgical quality
Numerical simulation has helped Ferroatlántica for ELSA
design and operation, and also for marketing

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Silicon

Silicon (Si) is the second most abundant element in the earth’s


crust after oxygen.
In natural form, it can be found mainly as silicon dioxide
(Silica, SiO2 ) and silicates.
In particular, quartz and sand are two of the most common
forms.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Applications of silicon

Depending on its purity, silicon has a wide variety of applications:


Ferrosilicon (silicon steels, it can contain more than 2% of
other materials)
Metallurgical silicon (e.g. silicon-aluminum alloys, it contains
about 1% of other elements)
Chemical silicon (silicones)
Solar silicon (solar cells)
Electronic silicon (semiconductors, the purest silicon, “9N” =
99.9999999 of purity)

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Metallurgy of silicon I

Silicon is produced industrially by reduction of silicon dioxide


with carbon by a reaction which can be written in a simple way
as follows:
Si O2 + 2C −→ Si + 2CO.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Metallurgy of silicon II

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

The ELSA electrode I


Motion system Graphite core

Nipple

Support
system

Pre-baked paste
Casing

Liquid paste

Clamps

Solid paste

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

The ELSA electrode II

Modulus of the current density, |Jh |, in conductors.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

The ELSA electrode III

e h in dielectric.
Magnetic potential Φ

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

The ELSA electrode IV

|Jh |: Horizontal section of one of the electrodes.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

The ELSA electrode V

|Jh |: Vertical section of one of the electrodes.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Induction furnaces

Photographs taken from http://www.ameritherm.com

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Industrial induction furnace

Silicon for
melting and
purification
Graphite
crucible
Refractory
layers
Water-cooled
coil

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Mathematical modelling

Multi-physics problem

Three coupled models corresponding to three different areas of


physics.

Electromagnetic model Thermal model

Hydrodynamic Model

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation I

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation II

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation III

VELOCITY

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation IV

VELOCITY

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electric motors

Contract with ORONA Company in the framework of the


project NET0LIFT to design new lift technologies
This project was co-financed by the Spanish research
programme CENIT
Our tasks were related to the numerical simulation of electric
motors

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Usage of electric motors I

Electric motors are a reliable way of transforming electrical energy


into movement.

Photo by Zurecks.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Usage of electric motors II

From classical uses. . . .

Photo by Harrihealey02.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Usage of electric motors III

. . . to present or future applications.

Photo by Tony Hisgett.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation of electric motors

Electromagnetic energy is dissipated into heat through two


different mechanisms: Joule effect and hysteresis
The released heat causes the temperature rise of the motor
This is a very important limiting factor in designing electric
motors.
Numerical simulation is nowadays an essential tool for
optimum design.

It is done in two steps:

1 electromagnetic analysis and

2 thermal analysis.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic analysis I

Electromagnetic analysis aims at determining the eddy currents


which are responsible for Joule heating.
Electromagnetic models are obtained from Maxwell equations
Three-dimensional electromagnetic simulation of the whole
motor is still a challenge

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic analysis II

Figure: Stator: coils and laminated core

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic analysis III

Figure: Magnetic laminated core

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic analysis IV

The presence of a laminated core composed of isolated thin


plates makes this problem difficult because very fine meshes
need to be used (several geometric scales)
The material of the laminate has nonlinear magnetic behaviour
with hysteresis
For these reasons, motor designers usually employ simplified
2D transient magnetic models.
These 2D distributed parameter models are coupled with
lumped parameter (or circuit) models

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Motor sketch

A motor is composed of many pieces:

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic simulation: 2D mesh

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic simulation: magnetic flux density

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Electromagnetic simulation: losses

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Heat transfer simulation. 3D FE mesh

Having many pieces of different size and local phenomena can lead
to large meshes . . .

A coarse mesh of the motor.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Temperature I

FEM GLPM

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Temperature II

FEM GLPM

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Microwave heating and food technology

Household microwave ovens work by passing non-ionizing


microwave radiation, at a frequency of 2.45 GHz.
Some substances in the food, like water and fat, absorb energy
from the microwaves in a process called dielectric heating
(lossy dielectrics).
Raytheon Company sold the first microwave oven in 1947,
derived from radar technology developed during the World War
II.
The compact versions become popular from 1967.
Microwave heating is very important for food technology, in
particular, for defrosting food

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Microwave oven

Figure: Domain.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Microwave oven

Figure: Domain.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation: electric field

Figure: Norm of the electric field.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation: temperature I

Figure: Temperature.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electrodes for electric arc furnaces
Electromagnetic modelling
Induction furnaces
Magnetic field formulation
Electric motors
Magnetic field/magnetic potential formulation
Microwave heating
Numerical solution

Numerical simulation: temperature II

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling Maxwell’s equations
Magnetic field formulation Constitutive laws
Magnetic field/magnetic potential formulation Harmonic eddy currents model
Numerical solution

Electromagnetic modelling: Maxwell’s equations

 ∂D

 ∂t − curl H = −J in R3 ,

 ∂B
Maxwell’s ∂t + curl E = 0 in R3 ,
equations 
 div B = 0 in R3


div D = ̺ in R3

H: Magnetic field, E: Electric field,


D: Electric displacement, J : Current density,
B: Magnetic induction, ̺: Charge density,
t: Time.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling Maxwell’s equations
Magnetic field formulation Constitutive laws
Magnetic field/magnetic potential formulation Harmonic eddy currents model
Numerical solution

Electromagnetic modelling: constitutive laws

J = σE + v × B, σ: electric conductivity
D = εE, ε: electric permittivity,
B = µH, µ: magnetic permeability.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling Maxwell’s equations
Magnetic field formulation Constitutive laws
Magnetic field/magnetic potential formulation Harmonic eddy currents model
Numerical solution

Harmonic regime. Eddy currents model I

(
∂D
∂t can be neglected (low frequency)
Assumptions
F(x, t) = Re [eiωt F(x)] (alternating current)

ω: angular frequency, i: imaginary unit

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling Maxwell’s equations
Magnetic field formulation Constitutive laws
Magnetic field/magnetic potential formulation Harmonic eddy currents model
Numerical solution

Harmonic regime. Eddy currents model II

Time harmonic eddy current model

curl H = J
curl E = −iωB
div B = 0
B = µH
J = σE

E(x) = O(|x|−1 ) uniformly for |x| → ∞


H(x) = O(|x|−1 ) uniformly for |x| → ∞
Fields are complex valued

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling Maxwell’s equations
Magnetic field formulation Constitutive laws
Magnetic field/magnetic potential formulation Harmonic eddy currents model
Numerical solution

Harmonic regime. Eddy currents model III

Different formulations are possible

Magnetic field/scalar magnetic potential (H/ϕ)


Magnetic vector potential/scalar electric potential (A/V )
Electric field (E)
Primitive of the electric field with respect to time (A∗ or u)
...

For the harmonic regime see, for instance, the book

Eddy Current Approximation of Maxwell Equations.


Theory, Algorithms and Applications
A. Alonso Rodrı́guez and A. Valli
Springer, 2010.
Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications
Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

The magnetic field formulation

Ω is a bounded domain such that


Ω = ΩC ∪ ΩD

ΩC : conductors (σ > 0)
ΩD : dielectrics (air) (σ = 0)
Conductors are not assumed to
be totally included in Ω

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Boundary conditions

The boundary of the domain splits as follows:


∂Ω = ΓC ∪ ΓD
where
ΓC := ∂ΩC ∩ ∂Ω,
ΓD := ∂ΩD ∩ ∂Ω.
ΓI := ∂ΩC ∩ ∂ΩD .
Boundary conditions:
E×n=g on ΓC,
H×n=f on ΓD.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Strong problem

SP.- To find the magnetic field H and the electric field E in Ω,


satisfying
curl H = 0 in ΩD , (1)
curl H = J = σE in ΩC , (2)
iωµH + curl E = 0 in Ω, (3)
div(µH) = 0 in Ω, (4)
E×n=g on ΓC, (5)
H×n = f on ΓD. (6)

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Weak formulation I

We eliminate the electric field E in the previous system:


Let us make the scalar product of (3) by a test field G such that
curl G = 0 in ΩD and G × n = 0 on ΓD .
Then, let us integrate in Ω. We get

Z Z
iωµH · Ḡ + curl E · Ḡ = 0 in Ω,
Ω Ω

Now we use a Green’s formula to transform the second integral:

Z Z Z
curl E · Ḡ = E · curl Ḡ + E · Ḡ × n dΣ.
Ω Ω Γ

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Weak formulation II

We have
Z Z
iωµH · Ḡ + curl E · Ḡ
0=
Z Z Z Ω Ω

= iωµH · Ḡ + E · curl Ḡ + E · Ḡ × ndΣ


Ω ΩC ΓC
Z Z Z
= iωµH · Ḡ + E · curl Ḡ + n × (E × n) · Ḡ × n, dΣ
Ω ΩC ΓC
Z Z Z
= iωµH · Ḡ + E · curl Ḡ + n × g · Ḡ × n dΣ
Ω ΩC ΓC
Z Z Z
= iωµH · Ḡ + E · curl Ḡ − g × n · Ḡ × n dΣ.
Ω ΩC ΓC

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Weak formulation III

Now we use (2) to deduce


1
E= curl H in ΩC .
σ
By replacing this equality we finally get

Z Z Z
1
iωµH · Ḡ + curl H · curl Ḡ = g × n · Ḡ × n dΣ
Ω ΩC σ ΓC

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Function spaces


G ∈ L2 (Ω)3 : div G ∈ L2 (Ω) ,
H(div, Ω) :=

H(curl, Ω) := G ∈ L2 (Ω)3 : curl G ∈ L2 (Ω)3 .

Hr (curl, Ω) := G ∈ Hr (Ω)3 : curl G ∈ Hr (Ω)3 , r > 0.
Each of these spaces is endowed with its natural norm, i.e.,
kGk2Hr (curl,Ω) = kGk2Hr (Ω)3 + k curl Gk2Hr (Ω)3 .

1/2
H00 (Γ): space of functions defined on Γ that extended by 0
to ∂Ω \ Γ belong to H1/2 (∂Ω).
−1/2 1/2
H00 (Γ): dual space of H00 (Γ).

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Weak formulation in terms of H

WP.- To find H ∈ V such that


H × n = f in H00 (ΓD)3 ,
−1/2
Z Z
1

iω µH · Ḡ + curl H · curl Ḡ = g × n, Ḡ × n Γ , ∀G ∈ V 0 .
Ω ΩC σ C

where
V = {G ∈ H(curl, Ω) : curl G = 0 in ΩD } ,
n o
V0 G ∈ V : G × n = 0 in H00 (ΓD)3 .
−1/2
=

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Existence of weak solution

Theorem.-If there exists Hf ∈ V such that Hf × n = f in


H00 (ΓD)3 , then the weak problem WP has a unique solution.
−1/2

The proof is standard (Lax-Milgram Lemma)

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Strong problem
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Properties of the weak solution

Theorem.- Let H ∈ V be the solution of problem WP. Let


B = µH ∈ L2 (Ω)3 , J = curl H ∈ L2 (Ω)3 , and
E = ( σ1 J)|Ω ∈ L2 (ΩC)3 . Then:
C

div B = 0 in Ω,
iωµH + curl E = 0 in ΩC ,
(ΓC)3 ,
−1/2
E×n = g in H00
in H00 (ΓD)3 ,
−1/2
H×n=f
J=0 in ΩD .
Remark: The electric field is not uniquely determined in the
dielectric domain

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Scalar magnetic potential
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Cutting surfaces

Bossavit and Verité (1982): magnetic field in ΩC / scalar


magnetic potential in ΩD .

Σj “cut”surface:
Σj ⊂ ΩD , j = 1, . . . , J.
∂Σj ⊂ ∂ΩD , j = 1, . . . , J.
Σ̄j ∩ Σ¯k = ∅ for j 6= k.
S
Ωe := ΩD \ j=J Σ̄j
D j=0
is simply connected.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Scalar magnetic potential
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

The kernel of the curl operator

e ) defined by
Let T be the linear space of H1 (Ω D
n o
T = Ψ e ∈ H1 (Ωe ) : [[Ψ]]
e
D Σj = constant, j = 1, . . . , J .

e ∈ T /C,
For all G ∈ V, there exists a unique scalar field Ψ
such that G|ΩD = gread Ψ.e

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Scalar magnetic potential
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Magnetic field/magnetic potential formulation

e ∈ W such that
Problem WPP: To find (H, Φ)
e × n = f in H−1/2 (Γ )3 ,
gread Φ 00 D
Z Z
1
iω µH · Ḡ + curl H · curl Ḡ
ΩC ΩC σ
Z


+iω e · gread Ψ
µ gread Φ ē = g × n, Ḡ × n , ∀(G, Ψ) e ∈ W 0.
Γ C
ΩD

where
n o
e ∈ H(curl, Ω ) × (T /C) : (G| gread Ψ)
W := (G, Ψ) e ∈ H(curl, Ω) ,
C
n o
e ∈ W : gread Ψ
W 0 := (G, Ψ) e × n = 0 in H−1/2 (Γ )3 .
00 D

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Electromagnetic modelling
Scalar magnetic potential
Magnetic field formulation
Weak formulation
Magnetic field/magnetic potential formulation
Numerical solution

Magnetic field/magnetic potential formulation

Theorem.- If there exists Hf ∈ V such that Hf × n = f in


e
H00 (ΓD)3 , then problem WPP has a unique solution (H, Φ),
−1/2

e
with (H| gread Φ) being the unique solution of problem WP.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Numerical solution: Finite Element Method

Ω, ΩC and ΩD are Lipschitz polyhedra.

{Th }: family of regular thetraedral meshes of Ω.

For every mesh Th , each element K ∈ Th is contained either in


ΩC or in ΩD.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Nédélec edge elements (Nédélec, 1980)

H(curl, Ω) is approximated by:


N h (Ω) := {Gh ∈ H(curl, Ω) : Gh |K ∈ N (K) ∀K ∈ Th } .
where

N (K) := Gh ∈ P1 (K)3 : Gh (x) = a × x + b, a, b ∈ C3 , x ∈ K .
Degrees of freedom of a function Gh ∈ N (K):
Z 
Gh · te for the six edges e of K .
e

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Nédélec edge elements. An important property

Fields of the form


Gh (x) = a × x + b, a, b ∈ C3 ,
have constant tangential component along any straight line in
the space
In particular, this is true along the 6 edges of any tetrahedron
This function vector space has dimension 6
The values of the tangential components are taken as
interpolation conditions to determine a and b in each
tetrahedron of the mesh

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Discretizing problem WP I

Problem DWP: To find Hh ∈ V h such that


Hh × n = fI on ΓD,
Z Z
1
iω µHh · Ḡh + curl Hh · curl Ḡh
Ω ΩC σ
Z
= g × n · Ḡh × n ∀Gh ∈ V 0h ,
ΓC

where
fI := two-dimensional Nédélec interpolant of n × f ,
V h := {Gh ∈ N h (Ω) : curl Gh = 0 on ΩD } ,
V 0h := {Gh ∈ V h : Gh × n = 0 on ΓD} .

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Discretizing problem WP II

Theorem.- Let us assume that the solution H of problem WP


satisfies H|ΩC ∈ Hr (curl, ΩC) and H|ΩD ∈ Hr (ΩD)3 , with
r ∈ ( 12 , 1].
Then, fI is well defined by the 2D Nédélec interpolant of n × f ,
problem DMP has a unique solution Hh , and
h i
kH − Hh kH(curl,Ω) ≤ Chr kHkHr (curl,Ω ) + kHkHr (Ω )3 .
C D

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Discretizing problem WPP I

Finite dimensional space to approximate T :


e h ∈ Lh (Ω
Th := {Ψ e ) : [[Ψ
e h ]] = constant, j = 1, . . . , J},
D Σj

being
n o
e ) := Ψ
Lh (Ω e h ∈ H1 (Ω e h | ∈ P1 (K) ∀K ∈ T ΩD .
e ): Ψ
D D K h

The curl-free vector fields in N h (ΩD ) admit a multivalued


potential in Th .

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Discretizing problem WPP II

e h ) ∈ W h such that
Problem DWPP: To find (Hh , Φ

gread Φe h × n = f on Γ ,
I D
Z Z
1
iω µHh · Ḡh + curl Hh · curl Ḡh
ΩC ΩC σ
Z
+iω µ gread Φe h · gread Ψ
ē h =
ΩD
Z
g × n · Ḡh × n ∀(Gh , Ψe h) ∈ W 0 ,
h
ΓC
n o
e h ) ∈ N h (ΩC ) × (Th /C) : (Gh | gread Ψ
W h := (Gh , Ψ e h ) ∈ H(curl, Ω) ,
n o
W 0h := (Gh , Ψe h ) ∈ W h : gread Ψ
e h × n = 0 on Γ .
D

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Discretizing problem WPP III

e of problem HP
Theorem.- Let us assume that the solution (H, Φ)
r e
satisfies H ∈ H (curl, ΩC) and gread Φ ∈ H (ΩD)3 , with
r

r ∈ ( 12 , 1].
Then, problem DWPP is well posed, it has a unique solution
(Hh , Φe h ), and

e − gread Φ
kH − Hh kH(curl,Ω ) + k gread Φ e hk 2
L (ΩD)3
h C i
≤ Chr kHkHr (curl,Ω ) + k gread Φke r
H (Ω ) .
3
C D

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Computer implementation of problem DWPP

It is necessary to impose the following constraints:


e h ) ∈ H(curl, Ω):
(Gh | gread Ψ
Elimination of the degrees of freedom of Gh on the interface by static
condensation.
e h ]] = constant, which arise from the definition of Th :
[[Ψ Σj

e h| − = Ψ
Ψ e h | + + [[Ψ
e h ]] = Ψ
e h | + + chj , j = 1, . . . , J.
Σ j Σ jΣj Σ j

e h × n = f on Γ is imposed
The boundary condition gread Φ I D
by means of a Lagrange multipler defined on ΓD.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Numerical experiments

Coaxial cable: Ω is a cylindrical domain which consists of two


different conductors ΩC1 and ΩC2 separated by a dielectric ΩD .
An alternating current J goes through the innermost
conductor along its axis.
Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications
Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Analytical solution

By using a cylindrical coordinate system:


In the innermost conductor: H(r) = c I1 (γ1 r)eθ .
I
In the dielectric domain: H(r) = 2πr eθ .
In the outer conductor: H(r) = (d I1 (γ3 r) + e K1 (γ3 r))eθ ,
being
√ √
γ1 = iωµσ1 , γ3 = iωµσ3
I1 and K1 modified Bessel functions.
c, d, e constants obtained by using the boundary and the
interface conditions.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Numerical results

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Numerical results

Error curve for the magnetic field H (log-log scale).


2
10
Percentual errors
y=Ch
Relative error (%)

1
10
4 5 6
10 10 10
Number of d.o.f.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

Numerical results

Magnetic potential in the dielectric domain.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

References

A. Bermúdez, D. Gómez, M.C. Muñiz, P. Salgado, R. Vázquez, Numerical


modelling of industrial induction, in Advances of Induction & Microwave
Heating of Mineral and Organic Materials p. 75–100, INTECH Open Access
Publisher. Rijeka. 2011.

A. Bermúdez, C. Reales, R. Rodrı́guez, P. Salgado, Numerical analysis of a


finite-element method for the axisymmetric eddy current model of an induction
furnace. IMA J. Numer. Anal., 30 p. 654–676, 2010.

A. Bermúdez, D. Gómez, P. Salgado, Eddy-current losses in laminated cores


and the computation of an equivalent conductivity. IEEE Transactions on
Magnetics, 44, p. 4730–4738, 2008.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications


Introduction. Industrial applications
Finite element approximation
Electromagnetic modelling
Discretized problems. Error analysis
Magnetic field formulation
Computer implementation
Magnetic field/magnetic potential formulation
Numerical experiments
Numerical solution

References

A. Bermúdez, R. Rodrı́guez, P. Salgado, A finite element method for the


magnetostatic problems in terms of scalar potentials. SIAM Journal on
Numerical Analysis, 46, p. 1338–1363, 2008.

A. Bermúdez, D. Gómez, M. C. Muñiz, P. Salgado, Transient numerical


simulation of a thermoelectrical problem in cylindrical induction heating
furnaces. Advances in Computational Mathematics, 26, p. 39–62, 2007.

A. Bermúdez, R. Rodrı́guez, P. Salgado, FEM for 3D eddy current problems in


bounded domains subject to realistic boundary conditions. An application to
metallurgical electrodes. Archives of Computational Methods in Engineering,
12 (1), p. 67–114, 2005.

A. Bermúdez, R. Rodrı́guez, P. Salgado, A finite element method with


Lagrange multipliers for low frequency harmonic Maxwell equations. SIAM
Journal on Numerical Analysis, 40 (5), p. 1823–1849, 2002.

Alfredo Bermúdez de Castro Numerical methods in electromagnetism and applications

Вам также может понравиться