Glass 1
Radiative Heat transfer and Applications for Glass Production Processes
Axel Klar and Norbert Siedow
Department of Mathematics, TU Kaiserslautern
Fraunhofer ITWM Abteilung Transport processes
Montecatini, 15. – 19. October 2008
Glass 2
Radiative Heat transfer and Applications for Glass Production Processes Planning of the Lectures
1. Models for fast radiative heat transfer simulation
2. Indirect Temperature Measurement of Hot Glasses
3. Parameter Identification Problems
Glass 3
Indirect Temperature Measurement of Hot Glasses
N. Siedow
Fraunhofer-Institute for Industrial Mathematics,
Kaiserslautern, Germany
Montecatini, 15. – 19. October 2008
Glass 4
Indirect Temperature Measurement of Hot GlassesOutline
1. Introduction
2. Some Basics of Inverse Problems
3. Spectral Remote Sensing
4. Reconstruction of the Initial Temperature
5. Impedance Tomography
6. Conclusions
Glass 5
Models for fast radiative heat transfer simulations 1. Introduction
Temperature is the most important parameter in all stages of glass production
Homogeneity of glass melt Drop temperature Thermal stress
To determine the temperature:
Measurement Simulation
Glass 6
Indirect Temperature Measurement of Hot Glasses 1. Introduction
With Radiation
Without Radiation
Temperature in °C
Conduct
ivit
y in W
/(K
m)
Radiation is for high
temperatures the dominant
process
Heat transfer on a
microscale
Heat radiation on a
macroscale
mm -
cm
nm
Glass 7
Indirect Temperature Measurement of Hot Glasses 1. Introduction
Heat transfer on a
microscale
Heat radiation on a
macroscale
mm -
cm
nm
( , ) ( ( ) ( , )) , ( , )m m t
Tc r t k r T r t r t D
t
)(Tqr
( , , ) ( ) ( , , ) ( ) ( ( , ), )I r I r B T r t
20
( , , )r
S
q I r d d
0( ,0) ( ),T r T r r D
+ boundary conditions
),())(1(),',()(),,( agg TBrIrI
Glass 8
Indirect Temperature Measurement of Hot Glasses 1. Introduction
Direct Measurement
• Thermocouples
Indirect Measurement
• Pyrometer(surface temperature)
• Spectral Remote Sensing
Glass 9
Indirect Temperature Measurement of Hot Glasses 1. Introduction
Glass is semitransparent
Inverse ProblemSpectrometerT(z)
[µm]10 2 3
0.8
4
1
0.6
0.4
0.2
0
Emissivity
[°C]
Depth [mm]10 2 3 4
1000
950
900
Spectral Remote Sensing
Glass 10
Inverse Problems are concerned with finding causes for an observed or a desired effect.
Control or Design, if one looks for a cause for an desired effect.
Identification or Reconstruction, if one looks for the cause for an observed effect.
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 11
Example 1:
Black Box
Input Signal
Output Signal - Measurement
I gf
0
( ) ( ) ( ) ( )x
Af x I x t f t dt g x
Assume: 1I 0
( ) ( ) ( )x
Af x f t dt g x If: • Continuous differentiable
(0) 0g •
Solution: ( ) '( )f x g x
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 12
Example 1:
21( )
2g x x ( )f x x
We find:Given is:
• analytically
• exact
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 13
Example 1:
21( )
2( )
g x x
f x x
2
4 5 4
1( ) ( )
2
( ) 10 sin(10 ) 10
g x x x
x x
A small error in the measurement causes a big error in the reconstruction!
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 14
Example 1: Numerical Differentiation
In praxis the measured data are finite and not smooth
1( ), , 1, 2,...,i i i i ig g x h x x i n ig
( )i if f x 1 1
2i i
i h i
g gf D g
h
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 15
Example 1: Numerical Differentiation
1%, 0.1h
1%, 0.01h
• A finer discretization leads to a bigger error
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 16
Example 2: Parameter Identification
0 1( ) ( ) ( ), 0 , (0) (0) , ( )u u
a x x f x x l a g u l gx x x
Practical meaning: Heat transfer equation
( )u x Temperature ( )a x Thermal conductivity
Diffusion equation
( )u x Concentration ( )a x Diffusivity
„Black-Scholes“ equation
( )u x Option price ( )a x Stock price
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 17
Example 2: Parameter Identification
0 1( ) ( ) ( ), 0 , (0) (0) , ( )u u
a x x f x x l a g u l gx x x
Practical meaning: Electrical potential equation
( )u x Electrical potential
( )a x Electrical conductivity
0
0
( )
( )
x
g f y dy
a xux
Knowing the potential find the conductivity
Elasticity equation
( )u x displacement ( )a x Youngs Modulus
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 18
Example 2: Parameter Identification
( ) ( ) 4 2, 0 1, (0) (0) 4, ( ) 4u u
a x x x x a u lx x x
0
0
( )
( )
x
g f y dy
a xux
( )u x
0%
Exact Measurement
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 19
( ) ( ) 4 2, 0 1, (0) (0) 4, ( ) 4u u
a x x x x a u lx x x
Example 2: Parameter Identification
0
0
( )
( )
x
g f y dy
a xux
( )u x
0.001% 0.001%
Noisy Measurement
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 20
( ) ( ) 4 2, 0 1, (0) (0) 4, ( ) 4u u
a x x x x a u lx x x
Example 2: Parameter Identification
0
0
( )
( )
x
g f y dy
a xux
( )u x
0.01%
Noisy Measurement
0.01%
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 21
Example 4:
1 1.99999801 62 1.99 10 8
3 1.0 10 0x 1x
1 8f 2 4f 3 0f
( ) ( ) ( ), 0 ,
(0) (0) 0, ( ) ( ) ( )
ux x f x x l
x x
u ul l u l
x x
1
2
2
1 1 0 1
1 1.000001 0.000001 1
0 0.000001 1.000001 0
u
u
u
0.01
0.01
0
Reconstruction:
1
2
2
20001.03
20000.02
0.000002
u
u
u
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 22
A common property of a vast majority of Inverse Problems is their ill-posedness
A mathematical problem is well-posed, if
Hadamard (1865-1963)
1. For all data, there exists a solution of the problem.
2. For all data, the solution is unique.
3. The solution depends continuously on the data.
A problem is ill-posed if one of these three conditions is violated.
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 23
What is the reason for the ill-posedness?
Example 1: ( ) '( )f x g x
( ) ( ) sin( )g x g x nx
4 510 , 10n
( ) ( ) cos( )f x f x n nx n
A small error in measurement causes a big error in reconstruction
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 24
What is the reason for the ill-posedness?
Example 1: ( )i h if x D g
2
''6h
hD g f f
h
optimalh
Step size must be taken with respect to the measurement error
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 25
What is the reason for the ill-posedness?
Example 2:
0
0
( )
( )
x
g f y dy
a xux
Numerical differentiation of noisy data
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 26
What is the reason for the ill-posedness?
Example 4:
1
2
3
1 1 0 1
1 1.000001 0.000001 1
0 0.000001 1.000001 0
u
u
u
Eigenvalues:
61 2 30.5 10 , 1.000001, 2.0000005
Condition number: 64 10
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 27
What is the reason for the ill-posedness?
Example 4:
1
2
3
1 1 0 1
1 1.000001 0.000001 1
0 0.000001 1.000001 0
u
u
u
Eigenvalues:
61 2 30.5 10 , 1.000001, 2.0000005
1 2 3, ,v v vLet be the eigenvectors
31
1
( , )i i ii
u f v v
The solution can be written as:
A small error in f f 3 3
1 1
1 1
( , ) ( , )i i i i i ii i
u f v v v v
2
61
1020000
0.5 10
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 28
What can be done to overcome the ill-posedness?
Regularization
31
6
0.5
( , ) 0.5
0.5 10i i iu f v v
2i
Regularization Methods
1. Truncated Singular Value Decomposition
We skip the small eigenvalue (singular values)
identical to the minimization problem
2( ) min
LJ u Au f
and take the solution with minimum norm
2min
Lu
Replace the ill-posed problem by a family of neighboring well-posed problems
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 29
Regularization Methods
2. Tichonov (Lavrentiev) Regularization
We look for a problem which is near by the original and well-posed
We increase the eigenvalues
3
1
1( , )i i
i i
u f v v
How to choose ?
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 30
2Lu
* 0.01977
2LAu f
Regularization Methods
2. Tichonov (Lavrentiev) Regularization
3
1
1( , )i i
i i
u f v v
*
8
1.0004
0.0102
1 10
u
Take 0n
L-curve method
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 31
Regularization Methods
3. Landweber Iteration
* *A Au A fWe consider the normal equation
1 * 0( ),k ku u A Au f u givenUse a fixed point iteration to solve
Iteration number plays as regularization parameter
1
k
Stopping rule = discrepancy principle
*( , ) inf |k f k N Au f
Solution after 4 iterations: * 70.5 0.5 1 10Tku
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 32
Regularization Methods
4. Classical Tichonov Regularization
* *A Au Iu A f Regularization of the normal equation
Equivalent to the minimization problem
2 2
2 2( , ) min
L LJ u Au f u
Tichonov (1906-1993)
Dealing with an ill-posed problem means to find the right balance between stability and accuracy
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 33
Regularization Methods
4. Classical Tichonov Regularization
* *A Au Iu A f Regularization of the normal equation
Equivalent to the minimization problem 2 2
2 2( , ) min
L LJ u Au f u
L-curve
7
0.4965
0.4965
5 10
u
7
0.5
0.5
5 10
u
Discrepancy principle
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 34
Regularization Methods
4. Classical Tichonov Regularization
* *A Au Iu A f Regularization of the normal equation
Equivalent to the minimization problem 2 2
2 2( , ) min
L LJ u Au f u
2 2
2 2( , ) 1 min
L LJ u Au f u
To get a better solution we need to include more information!
Assume:2
21
Lu
9
1.00104
0.00104
9 10
u
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 35
Indirect Temperature Measurement of Hot Glasses 3. Spectral Remote Sensing
formal solution:
s
s
D
aag
dsdseeB
TBeTBnee
sDs
DD
D
0
2
0
122
22
'),(},])[,(),(
)],(),([)1{(1
1
)'(sTT T
)(mI
Non-linear, ill-posed integral equation of 1. kind
One-dimensional radiative transfer equation
),()(),,()(),,( TBzIzz
I
Glass 36
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Formal solution:
s
s
D
aag
dsdseeB
TBeTBnee
sDs
DD
D
0
2
0
122
22
'),(},])[,(),(
)],(),([)1{(1
1
)'(sTT T
)(mI
mITG )(
Linearization: )]()[(*)'())((')(*)'( 1 km
kkkkk TGITGTTTGTG
)]()[(*)'()]()(')(*)'[( 1 km
kkkkk TGITGTTTGTG AkRegularization:
Glass 37
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
(Rosseland-Approximation)
0
)),,(()(
1
3
4
dtzT
T
Bkr),,( tz
z
Tkq rr
is FDA ofA
cqz rq
Temperature satisfies the radiative heat transfer equation
Radiative Flux
Iteratively regularized Gauss – Newton method:
)]())()((*)'[(])(')(*)'[( 11 skk
mkkkkk TTTGITGTGTGTT
Ak
k
How to choose ?
A?
Glass 38
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Discrepancy principle
Iteratively regularized Gauss – Newton method:
)]())()((*)'[(])(')(*)'[( 11 skk
mkkkkk TTTGITGTGTGTT
Ak
k
How to choose ?
k? 0lim,1,01
k
kk
kk r
Stopping rule for k?
?
1,0)()(
nkITFITF kn
II m
Glass 39
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Furnace Experiment
Furnace
Glass slab
Thermocouples
Glass 40
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Drop Temperature
Glass 41
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
The Improved Eddington-Barbier-Approximation
If we assume that
D
0
)()),(()()1()( dsesTBI sm
D
sDsag
Dm dseesTBTBneI0
22
2 }])[),(()(),()1){(1()(
Glass 42
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
The Improved Eddington-Barbier-Approximation
If we assume that
D
0
)()),(()()1()( dsesTBI sm
)1)(1)(1(
),()1()()),((
)()(
22)(2
DD
agDm
ee
TBneIzTB
)1)(1)((
)1()()1)(1()()(
)()()(
DD
DDD
ee
eDeez
2/0 Dz
D
sDsag
Dm dseesTBTBneI0
22
2 }])[),(()(),()1){(1()(
Glass 43
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
A fast iterative solution of the integral equation
i. For :0l2
0),()(0 DzzTzT E Calculate using the IEB-method
iii. For :1ll Use to calculate
)(zT l
ii. Using some additional information continue to )(0 zT Dz 0
D
sDslag
Dl dseesTBTBneI0
22
2 }])[),(()(),()1){(1()(
Glass 44
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
A fast iterative solution of the integral equation
v. Calculate a new temperature profile in the IEB points using
)(1 zT l
20
Dz
)1)(1)(1(
)()()),(()),((
)()(1
DD
lmll
ee
IIzTBzTB
vi. Using the additional information continue to )(1 zT l Dz 0
and go back iii.
iv. If then STOP else continue with v. )()( lm II
Glass 45
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Iteration 1
Glass 46
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Iteration 2
Glass 47
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Iteration 3
Glass 48
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Iteration 4
Glass 49
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Iteration 5
Glass 50
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Iteration 10
Glass 51
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Error 1%
Glass 52
Indirect Temperature Measurement of Hot Glasses 3. Indirect temperature measurement
Error 1%
Glass 53
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
Initial condition
Boundary condition
Temperature
+ Additional Measurements of boundary temperature / heat flux
Is it possible to reconstruct the initial temperature distribution from boundary measurements?
?
Glass 54
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
*
0
0 (0, *) (0, )
(0, ) (0, )
( ,0) 0 (0, *)
( , ) 0 (0, *)
t xx t
x
x
u u in Q t
u x u on
u t t t
u t t t
Additional measurement at boundary
0Au y
( ,0)u t y
Problem is ill-posed:
• Solution exists
• Solution is unique
• Continuous dependence on right hand side is violated
Glass 55
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
Backward Heat Sideways Heat New Heat Problem
Given:
2 conditions at y2 Additional bc at yTemperature at T
Looking for:
bc at y1 Initial conditionInitial condition
Glass 56
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
2
00
( , ) ( , ) ( )n tn n
n
u t x u w e w x
0Au y
Problem is ill-posed:
• Solution exists
• Solution is unique
• Continuous dependence on right hand side is violated
1 0( )
2 cos( ) 0n
nw x
nx n
For2
( ) ( ) 2k k ty t y t k e
One obtains 0 0 2 cos( )ku u k kx
Glass 57
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
2
00
( , ) ( , ) ( )n tn n
n
u t x u w e w x
0Au y
Problem is ill-posed:
• Solution exists
• Solution is unique
• Continuous dependence on right hand side is violated
1 0( )
2 cos( ) 0n
nw x
nx n
For ( ) ( ) 0ky t y t
One obtains 0 0ku u
Glass 58
• Given data
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
hy Y
0Au y
Solve the system by Tichonov regularization
• While searching for the correct
Solve * *0,h h hA A I u A y
Adjust according to the parameter choice rule
with
hYy y
by some method
• is the regularized solution 0,1
N
n nn
u a u
Glass 59
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
*
0
0 (0, ) (0, )
(0, ) (0, )
( ,0) 0 (0, )
( , ) 0 (0, )
t xx t
x
x
u u in Q T
u x u on
u t t T
u t t T
L. Justen. An Inverse Heat Conduction Problem with Unknown Initial Condition. Diploma Thesis, TU KL, 2002
9( ) ( ,0) ty t u t e
Exact solution:
0 ( ) cos(3 )u x x
Tichonov regularization with Morozov‘s Discrepancy stopping rule
Glass 60
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
2D Example: Parabolic Profile – Four-Sided Measurement
Glass 61
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
Glass 62
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
S.S. Pereverzyev, R. Pinnau, N. Siedow. Proceedings of 5th Conf. on Inverse Problems in Engineering, 2005
S.S. Pereverzyev, R. Pinnau, N. Siedow. Inverse Problems, 22 (2006), 1-22
Initial Temperature Reconstruction for a Nonlinear Heat Equation:
Application to Radiative Heat Transfer PhD S.S. Pereverzyev
124
21
( , ) ( ), ( ) 4 2
( , ) ( ( , )),
( ,0) ( )
b b
T Tx t D T D T T Id
t x
Tx t B T x t
nT x u x
( ) : ( )D u D T : ( , ( ))Fu L u D u y
Glass 63
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
S.S. Pereverzyev, R. Pinnau, N. Siedow. Proceedings of 5th Conf. on Inverse Problems in Engineering, 2005
S.S. Pereverzyev, R. Pinnau, N. Siedow. Inverse Problems, 22 (2006), 1-22
: ( , ( )) ( ,0) (0, ( )) :Fu L u D u L u L D u Au Gu y
Decomposition of the non-linear equation:
Linear part non-linear
measurement
Fixed-point iteration: 1k kAu y Gu
Tichonov regularization:
* *1k kA A I u A y Gu
Glass 64
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
Parameter choice rule: Quasi-optimality criterion
• It does not depend on the noise level
1. Select a finite number of regularization parameters
0 10 ... m
which are part of a geometric sequence, i.e.
0 , 1ii q q
2. For each solve and obtaini * *1,( ) ( )
ii k kA A I u A y Gu 1, iku
3. Among 1, 0i
m
k iu
choose 1, jku such that
21 11, 1, 1, 1, (0, )min , 1,2,...,
j j i ik k k k L lu u u u i m
Glass 65
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
Glass 66
Indirect Temperature Measurement of Hot Glasses 4. Reconstruction of Initial Condition
Reconstruction is dependent on
• noise level
• discretization
How to choice the right discretization parameters depending on the noise level?
Glass 67
Indirect Temperature Measurement of Hot Glasses 5. Impedance Tomography
The knowledge of the temperature of the glass melt is important to control the homogeneity of the glass
Glass melting in a glass tank
Glass 68
Indirect Temperature Measurement of Hot Glasses 5. Impedance Tomography
• Thermocouples at the bottom and the sides of the furnace
• Use of pyrometers is limited due to the atmosphere above the glass melt
Glass 69
Glass melt
Determine the temperature of the glass melt during the melting process using an impedance tomography approach
23.2))(lg(890353)(
xxT
( )x
applyElectric current
measure Voltage
Neutral wire
Experiment
electrode
Indirect Temperature Measurement of Hot Glasses 5. Impedance Tomography
Glass 70
Inverse Problems are concerned with finding causes for an observed or a desired effect.
A common property of a vast majority of Inverse Problems is their ill-posedness(Existence, Uniqueness, Stability)
To solve an ill-posed problem one has to use regularization techniques(Replace the ill-posed problem by a family of neighboring well-posed problems)
The regularization has to be taken in accordance with the problem one wants to solve
Indirect Temperature Measurement of Hot Glasses 6. Conclusions
Glass 71
Further Examples of Inverse Problems:
Computerized Tomography
0
ln ( , )n
L
R
Idt s w
I ( , , )f s w t
Inverse Scattering
2 2 ( ),s s i su k u k u u f 21 ( ),f n x
2
( ) ( ) ,4
ik x ys ik e
u x u y dy x Rx y
( )f y
0I
LI
iusu
( ) 1/ ( )n x c x
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 72
Example 3:
1.0 0x 1x
1f 2f 3f
0 1
( ) ( ) ( ), 0 ,
(0) (0) , ( ) ( )
ux x f x x l
x x
u ug l l g
x x
1
0
(1) (1) (0) (0) ( ) 0u u
f x dxx x
0 1
0
( ) ,l
f x g g If exists no solution
0 1
0
( ) ,l
f x g g If exist an infinite number of solutions
Gauge condition
1
1 0
0
( ) 0g g f x dx
Integration
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 73
Example 4:
1 1.99999801 62 1.99 10 8
3 1.0 10 0x 1x
1 8f 2 4f 3 0f
( ) ( ) ( ), 0 ,
(0) (0) 0, ( ) ( ) ( )
ux x f x x l
x x
u ul l u l
x x
1
2
2
1 1 0 1
1 1.000001 0.000001 1
0 0.000001 1.000001 0
u
u
u
Exact Solution:
1
2
2
1
0
0
u
u
u
1
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems
Glass 74
What is the reason for the ill-posedness?
Example 3:
1 0 12
2 2
3 1 3
1 1 0
1 2 1 0 22
0 1 1
u g fh
u h f
u g f
System is singular
1 0 12
2 0 1 2
3 0 1 1 2 3
1 1 0
0 1 1 22
0 0 0 2
u g fh
u h g f f
u g g f f f
Indirect Temperature Measurement of Hot Glasses 2. Some Basics of Inverse Problems