Page 276 - Process Modelling and Simulation With Finite Element Methods

Coupling Variables Revisited 263

capable of reproducing nearly the same boundary data as the forward problem,
the boundary data is insufficient to select among the potential interior
configurations.
Omds I 01 El -
: : I id E?. ......
0 O m3s n om - E3 .... -
... t... ;::....... ...............
,?,
; ;;;';.:< ...............
>
0 01
0 O?
..
:rvp ................... 1
-..-.
.
....
--,.:-.
-
,-:
-.
001
/;:,
._,;.
-.,
0
I.
I.
401 I , I . , I
0 '0 40 . M 90 LW 110 14
iterate number
Figure 7.5 Optimization of distance from measured charges 91. q2, q3 by predictions of same with
dielectric constants from &I, ~2, ~3 chosen by Nelder-Mead optimization iterates. Left: error norm
(see m-file errornmm) with increasing iterate number; Right: dielectric constants versus iterate
number. Initial dielectric vector is [O.Ol,0.0l,O.Ol].
120 L4 0 '0 do , 60 En lm 110 ld
............... iterate number
Figure 7.6 Optimization of distance from measured charges 41. 42, q3 by predictions of same with
dielectric constants &I, &2, &3 chosen by Nelder-Mead optimization iterates. Left: error norm (see
m-file err0rnm.m) with increasing iterate number: Right: dielectric constants versus iterate number.
Initial dielectric vector is [0.1,0.1,0.1].
c
ow
; om
a
u _- oms
3 om
OM5
1.1 " 1
O J ~ " ' 1 '
0 10 20 10 do 50 M 70 91
iterate number
Figure 7.7 Optimization of distance from measured charges 91, q2, q3 by predictions of same with
dielectric constants &I, E~, &3 chosen by Nelder-Mead optimization iterates. Left: error norm (see
m-file errornmm) with increasing iterate number; Right: dielectric constants versus iterate number.
Initial dielectric vector is [0.O6,0.05,0.O4].