DAI
HOC
QUÓC GIA
HA
NOI
TRirÓNG
DAI HOC KHOA HOC TU
NHIÉN
A>
^"^
VNG
DUNG PHAN MEM MATHEMATICA
TRONG
GIÁNG
DAY VÁ NGHIÉN
CÚt
VAT LY.
MÁ SO:
QT-01-31.
CHÚ
TRÍ
DÉ
TAI: GS.TS.
TON
TÍCH
ÁI.
CÁC
CAN BÓ
THAM GIA:
- CN NGUYEN HOÁNG OANH.
- CN NGUYÉN VÁN
NGHIA.
HA NOI 2003
DAI
HQC
QUÓC GIA HA NOI
TRÜÓNG
DAI HOC KHOA HOC
TU"
NHIÉN
BAO CAO
K\ÍT
QUÁ
THUC
HIÉN
DE
TAI
KHOA HOC CÓNG NGHÉ
QG,
QT,
TN
1.
Ten
dé
tai:
Ung dung phan
niém Mathematica
trong giáng day vá
nghién
cxiu
vat ly.
Má so:
QT-01-31.
2.
Chú
tri
dé
tai:
GS.TS. Ton Tích Ái.
3.
Các
can
bo tham gia:
CN Nguyén Hoáng Oanh.
- CN Nguyen Van
Nghía.
4.
Muc tiéu
vá
noi dung nghién
cxiix:
Nám vüng he thó'ng dai
so
máy tính (CAS) Mathematica.
M5
hinh hoá
mgt só'bái toan
Vat ly phuc vu cho viec giáng
d.ay
các mon
vat ly dai
cu'dng.
Dé xuá't
du'dc
món nhán
bien
doi cho
trUdng
Dia Vat ly
dé thé
hien tính
toan
báng Mathematica
Tho'ng
nhá't các phép bié'n doi
trUdng
thé trong mién tan so' la cd
stJ
cho
các bien doi báng
ngón ngCí
dai
so'máy
tính (CAS).
Lap trinh báng ngon
ngií
Mathematica vé tá't
cá
các phép bien doi
trUdng
phuc vu cho viec
minh
giái các
só'heu dja
vat ly.
5.
Tinh hinh
thiic
bien dé (ái:
- Cong trinh dé cap den
mói
phUdng
huóng
nghién
cúu
mói, sü dung
niót
phán
mém toan
hoc mói
ti
ong giáng day vá
nghién
cúu vát ly, cu l,hé
lá
giáng day vat
ly
dai
cUóní
cüng nhu nghién cúu các
phu'dng phá])
phán
chia
trucíng
dia vát ly.
- Thánh
hip
các file phuc
vii
cho viec minh chúng mot
so'bái
giáng vé
val
ly dai
cUdng.
Các
chiídng ti hih dUdc
viét báng phán mém Mathematica.
- Mol file
chudng
trinh
dé tlióng nhát
các phép
bien
doi
ti'iícJng
\y:\n¡;
JIÍÍÓM
ngCí Mathematica.
- Thé hién
phUdng
pháp phan chia
trUdng
theo
phUdng
pháp yéu tó'chính
báng ngón ngü Mathematica.
6. Dánh giá chung.
Dé tai
dUdc thUc
hien có chá't
ludng
vói khó'i
lUdng
khá
lón.
Các két quá thu
dUdc
da dUdc
bao cao
tai các hoi
nghi
khoa hoc vé vát ly,
toan
tính
toan
cüng
nhU
dia vat ly trong
nUóc
vá quó'c té:
Da dáng trong Proceeding
hói
nghi khoa hoc
lán thü
15 cüa
tru'dng
dai hoc
mó dia chá't "Phán chia
triícíng
thé' báng he thó'ng dai so' máy tính (CAS)
Mathematica.
Da
bao cao
tai các hoi nghi:
School and 2-nd Workshop on Simulation and Modeling Physics. Institute
of Physics, tháng 11 nám 2002.
Bao cao
tai hoi nghi khoa hoc
Lhú
15 trUdng dai hoc
mó
dia chá't vá hoi nghi
khoa hoc lán thú ba trUdng dai hoc
klioa
hoc
tU
nhién. Phán chia
trUdng thé
báng he
tho'ng dai
so'máy
tính (CAS) Mathemal
ica.
TrUdng DH
Mó
Dia chá't.
Tu>
én táp
bao cao
hoi nghi khoa hoc
lán
thú 15.
Quyén 4. Dau khí. Há noi
15/11/200L1
Bao cao
tai hoi
toan
hoc tính
toan
quó'c té tai Há noi.
Mathematica
for
Geophysical
h'ield
Separation.International
Conference on
High Performance Scientific
<'omputing.
Modeling, Sihnulation and
GptimizationlDf
Complex Process.
M;irch
10-14,
2003 Hanoi, Vietnam.
Sü dung Multimedia (da
phUdnj^
lien) trong doi mói phu'dng ])háp giáng day vá hoc
vat ly.
Bo giáo due vá dáo tao. Hoi
liiáo
doi mói phUdng pháp
giáng
day vá hoc d
dai hoc, Cao
dáng.
Há noi tháng 3
nam
2003.
Sách chuyén kháo da xuát bán:
Dang in cuó'n " Phán mém
toáii
hoc cho ky sU ( Mathematica)
Két
quá
dáo tao:
Da dáo tao hai cü nhán da tó't nghiép
Két quá da
diíg^c iVng
dung:
Da dUa váo thu'c té giáng day
vi\t
ly.
Có nhiéu khá náng úng dung
i-ho
viec phán tích các
só'Iiéu
dia vát ly
Da hoán thánh dúng tién do
\a
các
chi
tiéu má
nhiém
vu dé ra.
6. Tinh hinh sü dung kinh
p!ií
cüa dé
tai.
Toan bo kinh phí: Tám triéu
dóng
( 8.000.000
dóng).
So'tién
náy dUdc dúng
cho tá't cá các
khoán
thué muón de
thuc
hien dé tai, thu thap
tai liéu
, in
á'n,
kinh phí bao ve dé
tai
cüng nhU nop le phí quan ly hánh chính cüa
Khoa Vat ly cüng nhU cüa trudng dai hoc Khoa hoc Tu nhién.
Há noi, ngáy 27 tháng 3 nám 2003
Xác nhán cüa han chü nhiém khoa
'^
Chü nhiém dé
t|ai
GS.TS.
Ton Tích
Ái
^1
cüa
cd
quan chü tri
PHÓHI^UinüÓNCi
VIETNAM NATIONAL
UNIVEkSlTY
COLLEGE OF NATURAL SCIENCES
REPORT
0^
THE
SITUATION
AND THE
RESUl
TS OF
TIIE PROJECT
1.
Title
of projcct: Mathematica in
invcstigating
and
teacliiiig
physics.
Code:
QT-Ol-31.
2.
Director
of projcct:
Prof.
Dv.
I
nn Tich Ai
3.
Participants:
B.Se
Nguyen
Ho;ingOanh,
B.Se
Nguyen
Van Nghia.
4.
Aiiii
and contcnt of
rcsearchcs:
The project contains differcnt
Icchniques
of geophysical data
pmccssing
by
usíng
software Mathematica.
It
was
knowii lli;il geophysical
data
are usual ly distortcd
due to thc
inlluence
of
noisc.
Predominantly,
the noisc can be caused by
gcological
inhomogeneous
structurc, by
ermrs
in measurements and
inclhods
being
used,
by
variations
of physical
ficlds,
etc.
As a malter of fact,
geophysical
dala proccssing is the combat with the noisc of various
natures. In geophysical data processing, thc
llcld
separation acquires
grcat
signiílcancc.
Field
separation is ene of thc
niost
important
problcnis
in
gcoph)sieíil
dala
interprelation. For potential
fíclds,
when
iliere
are observational data for
ihc bolh ihe
proílle
survey and arca survey. The field
separatlun
bccomcs a
process
ofestimation of
low
íVec|ucncy
component, i.e., the regional anomaly, on the one hand, and high- frequency
lleld
componcnt,
i.e. thc residual or local anomaly on the
oilier
hand.
Author in frequency
domain estal")Iislics
thc kernel of
Saxíiv
Nigarrd
'Iransform,
and
this method is sampled for comparison
\\Ílh othcr
methods. The
results
of
sainpling
allovv us
to use the method in practice.
In this project the
ficId separation
is rcalizcd in
frequency
domain by using Computer
Algebra
system-
Malheniatica.
In
geophysical data proccssing.
Iwo
problems
acquií'c
grcat signiílcancc: lleld
separation
and wcak signal detcction against
anibient
noisc.
Probleni
of
weak geopliysical
is
most
important
bccausc of thc
ncccss¡l\
for dclection of weak conlrasl
geological
objecl
which, ¡n turn, ¡s connected with dccp
dcposit
prospecling,
seismic
stratigraph_\'.
undcrground
cavitics and ground water prospeeting.
cW.
Apart
íroin
the known
information
on spectral
anaivsis.
on eorrelation functions and
processing.
inckiding ihe
lleld separation with the aid of
inain componcnt
análisis.
The rapport
also dcals
with thc using
nuiltimedia
in
Icaching
physics.
The results of
rcsearchcr
is uscd
in geopiíysical
lleld
separation
and
teaching plnsics
in
practice.
5.
Cunclusiun:
the project is
vvell
rcaiized. The
results
of investigation give ability of
using proposed methods in gravity and magnetic
prospeeting
oil,
gas and othcrs minerals. Thc
received results have been published by
iwo
reports at international conferences o physics and
mathematics instilules of center for natural seiences (2002, 2003).
The project has played a
part
in the training physics eourse of
gradLiated
students.
6. The using of Fund.
Total fund is
allowed
and obtained:
6,000,000.00
VND
For:
Realization of project: 7,000,000.00 VND
Printingand
fee:
UOOCOOO.OOVND.
Thc Director of Projcct
Prof.
Dr. Ton Tich Ai
PHAN I
PHÁN CHIA
TRÜÓNG
THE BÁNG HE THONG
DAI
SO MÁY TÍNH
(CAS),
MATHEMATICA
l.l.MÓDÁU.
Mo phóng
báng
so' các quá trinh vá hien tUdng ky thuát vá tu nhién khác
nhau tren các máy tính dien tü da
tro
nén mot cóng cu manh dé nghién cúu các
quá trinh dó. Trong dia vat ly
tháni
dó viec mó
phóng
báng só'dUa tren các phán
mém khác nhau nhám tu
chính
vá phán tách các so' liéu dia vát ly. Trong nhiéu
trUdng hdp viec xác dinh các thóng só' cüa các vát thé gáy nén các di
thUdng
dia
vát ly lá quá trinh chü yéu trong
viéc
nghién cúu các so' liéu dia vát ly
thUe
té.
Phán tách trudng lá mot trong nhüng quá trinh quan trong nhá't trong giai
doan minh giái các só' liéu dia vát ly. Vói các trUdng thé, khi có các so' liéu theo
tuyén cüng nhU theo dien, thi viec
]jhán
chia trUdng trd thánh quá trinh dánh giá
các thánh phán
tan
só'thá'p, túc lá các di thUdng khu vUc hoac các thánh phán
tan
só'cao lién quan den các di thUdng
dii
hoac di thUdng dia phUdng.
Dé thUc hien các ky thuát phán chia trUdng trong thUc té ngUdi ta can có
nhüng giá thuyé't nhá't dinh vé pho vé các cüa so loc,v.v
1.2. CÁC
PHÜONG
PHÁP
BIEN
DOI
TRÜÓNG
THE.
Muc dích chính cüa quá trinh phán chia các só' liéu dia vát ly, lá tú các só'
liéu quan sát tách ra các thánh phán lién quan vói các vát thé dia chá't nám tai
các do sáu khác nhau. Hám thu
dude
phu thuoe váo các
toan
tü bien dói có thé có
cüng ddn vi vói hám xuá't phát hay lá các dao hám cüa chúng. Các dao hám có thé
dUdc tính d múc xuá't phát hay d
niot
do sáu
nao
dó. Hám dude
bien
dói cüng có
thé có thú nguyén cüa tích toa do vói hám xuát phát.
TrUóc nhüng nám 80 do trinh do ky thuát
con
thap vá các cóng cu tính
toan
nghéo nán, da só'các bái
toan bien
dói trUdng phán lón
dUde
thUc hien trong mién
khóng gian. Hien tai theo khuynh
huóng
chung, các bái
toan
náy dUdc thuc hien
trong mién
tan
só'nhd các thuát
toan bien
dói Fourier hoac các phán mém khác.
1.3. BIEN DÓI TRONG MIÉN KHÓNG GIAN.
Biéu thúc
toan
hoc cüa
phéji
bien dói trUdng thóng dung trong mién khóng
gian dUdc thé hien nhU sau [2]:
- Trung binh hoá trong
i)hani
vi vóng trón bán
kinh
R:
1
¿n
n
y(0,0,0)
= y ^jV(r.a,0)r(/n¡a
- Tiép tuc giái tích trUdng den múc z
(1.1)
2rr
«
Kraa-r;=f
Jf^^"'?;r!^
(i-2)
27r J J
(r'-^z'j-
-
- Tính dao hám theo truc x
VJO.O z) ^±r(y(' a.O)=rco.sa<¡nla
27t¡l (r'-^-Jf
- Dao hám theo truc z
K„ra^ z;
=
f]]^^-' ;^f-'^-f>'^'^"
(1.4)
Trong trUdng hdp dac biet, dé thUc hien các
bien
dói ké
tren
ngUdi ta da
thié't ké' các
palet
khác nhau dé tién hánh các tính
toan
báng tay. Tuy váy,
các biéu thúc giái tích vúa néu cüng có thé sü dung dé thUc hien các phép
bien
dói
tren
các máy tính
dién
tü.
1.4.BIEN
DÓI TRONG
MÍEN
TAN SO.
Trong mién náy tá't cá các phép
bien
dói dUdc néu
tren
có thé
áxióc
biéu dién
theo cúng mot cóng thúc chung:
- Vói các bái
toan
3D:
VJx,y.z)=\\V(^jj,0)K(^ y.n-y,z)c¡^c¡n
(1.5)
s
- Vói eáe bái
toan
2D:
VJx,z)=¡V(^.0)K(^-x,z)c¡^
(1.6)
/.
trong dó
V^/x,y,z),V^/x.z)lk
các trUdng da dUdc bien
áoi.V^/i^jj.z),V^/(^,z)
lá
trUdng xuá't phát,
K(^-xjj-y,z).
!\(^-x,z)-
nhán
bien
dói tUdng úng cho bái
toan
3D vá 2D.
Các biéu thúc (5) vá (6) lá các tích phán cháp. Vé mat
toan
hoc, các quá
trinh loe déu du'dc mó tá báng các tích phán dang náy. Nhu váy các bái
toan
bien
doi trUdng có thé dUdc xem nhU lá các phép loe
tan
só' má trong dó các
toan
tü bien
doi khác nhau dUdc thUc hien bdi
VAC
dac trUng
tan
só'khác nhau.
Theo
ly
thuyé't tích phán cháp trong mién
tan
só',
ph6s,/co)cixa
hám
y,/x,y,z)hoac
yjx,z)\k
tích cüa pho cüa
V(i^./].()),(V(^,0))va
cüa
K(4-x.7],y,zl(K(^-x,z)),tÚclk:
SJcü)
=
S(co)0(co)
(1.7)
0(co) SJo))/S(ú))
(1.8)
trong dó
S(Q))\k
pho cüa
V(^,j],(}),(y(^,0)),0(cú)\k
pho ( dac trUng
tan
só) cüa
hkmK(4~x,i],y,z),(K(^-x,z)).
Sü dung bien doi Fourier trUe tiép,
5,/¿y^
trong cóng thúc (7) du'dc tính
toan,
vá sau dó sü dung bien dói Fourier
ngUde,
hám
VJx,y,z)á\iúQ
xác dinh.
Báng 1 sau dáy cho ta các dac tru'ng
tan
só'cüa các phép bien doi các so'lieu trong
lUc
vá tü khác nhau.
Báng 1
Phép bien dói cüng dac triíng
tan
só
Phép
bien
dói Dac trUng
tan
só'
-Trung binh hoá
Bái
toan
3D
2J¡(cor)/Sin(o)r)
Bái
toan
2D
sin(cor
)/((or)
- Tiép tuc giái tích
• Tiép tue lén tren
Exp(-coz)
• Tié'p tuc xuó'ng duói Exp((oz)
Dao hám tháng dúng bác n
• Tren mat quan sát co"
• Tai do cao z
co"
exp(-cúz)
- Dao hám nám ngang bac n
•
Tren
mat quan sát
(i(o)"
• Tai do cao z
(¡^j->
exp(-coz)
Dac triíng tan so cüa phép bien dói Sacxov- Nigarrd
Vé nguyén ly, phép
bien á(n
náy do Andreev vá
Grifin
(
Lién
xó cü) dé ra
trong nhüng nám 60 vói muc dích xác dinh các thánh phán bien
dói.
Sau náy, y
tuóng
náy
dUdc
các nhá dia vát ly phUdng Táy phát trien vá hoán thién. PhUdng
pháp
dude
dúng dé phát hien các thánh phán trUdng nám tai các dó sáu xác dinh.
Trong mién khóng gian,
phUdng
pháp náy
dUde
mó tá báng biéu thúc sau:
yjO,0,0)=^^'''^
^^'''^
(1.9)
trong dó y(r¡)vk
Vír^)
tUdng úng lá các dai
lUdng
trung binh cua trUdng quan sát
dUdc tren các vóng trón bán
kinh
j-,
vá
r^
Dac trUng tan só'cüa phép bien dói (9) dUdc xác dinh nhU sau:
Tú giáo trinh
toan
hoc, ta có cóng thúc:
__-?__= re-'"'JJú}r)cüilco
(1.10)
Vi váy cóng thúc Poisson
trong he thó'ng toa do tru sé
lá:
y(0.0,-z)
=
—
jj^V(r,a.O)e~"''JJü)r)rú}cIrcIcoda
^^
ü
u
o
trong dó
JJo)r)\k
hám Bessel
loai
mot hang khóng.
(l.ll)
Néu ta dUa váo giá tri trung binh cüa hám
V(r)
tren vóng trón bán
kinh
r
báng cóng thúc:
y(r)
=
—
\V(r,a,0)da
(1.12)
2n
i
thi có thé vié't
lai
cóng thúc (11) duói dang:
y(Q.O-z)^
^^(oe-'''dcú^yi7)JJcor)rdr
(1.13)
Khi
z=0
en
CD
y
(0,0,0)=
\(odco\V(r)JJ(or)nlr
(1.14)
TUdng tu vói cóng thúc
(l.'i),
ta có thé viét lai biéu thúc tích phán Fourier
cho giá tri trung binh cüa hám só
^''(rj tren
vóng trón bán
kinh
r,
V(rJ = rcúJJcoi] )dcú^V(r U„(cor)rdr (1.15)
Né'u ta dUa váo biéu thúc:
S(ü)) = ^ y(r)J,(Q)r)rdr (1.16)
thi các cong thúc (13) vá (15) trd t hánh:
y(0,0,'z)=^e-"''S(Q))codco
(1.17)
y(r,)= ^S(co)J,(cor,)Q)dco (1.18)
Nhu da dé cap
ó
tren,
phudng
pháp Saxov-Nigarrd trong mién khóng gian
(9) có thé dude viét lai duói dang sau:
rS{Q))[j,(cor,) -JJcor^)]
yjO,0,0)
=
-^ -codeo (1.19)
^2-n
Do dó, nhán bié'n doi (9)
lá
0(co)
=
e''"'va
nhán cüa phép bien dói Saxov-
Nigard lá:
^(^)^¿Ál!^]±:M^]ll
(1.20)
1.5. BIEN DÓI TRÜÓNG TRONG
LÜC
BÁNG PHÜONG PHÁP SAXOV-
NIGARD TRONG
MÍEN
TAN SO.
Dói tUdng cüa phUdng pháp Saxov-Nigarrd lá nghién cúu cau
truc
cüa vüng
nghién cúu tai các chiéu sáu
kh;'ic
nhau báng cách bié'n doi các tham só'
r,
vá
r^
trong cóng thúc (20)
Tü các phUdng trinh (20) vá (7), trong
thUe
té', dé thUc hien phép bien doi
Saxov-Nigarrd
can
phái sü
á\xn\:
algorithm sau:
(1) Xáy dUng ma trán
só'liOu
xuá't phát
n
chiéu
dulü(nji)
(2) Bié'n doi Fourier thuan
ma
trán
daía(n,njáé
có
dUdc
ma
trán pho
Fourier S(n,n)
(3) Theo cóng thúc (20), tính ma trán nhán dac trUng
tan só
0(n.nj.
Vói các
bié'n doi khác
,
0(n,n)
tính
dUoc lú
báng dac trUng
tan só'trong
báng 1.
(4) Tim ma trán
tan só'cüa
trUdng dá dUde
bien
doi
5',//;,/;^báng
cách tính
tích S(n.n)by 0(n,n).
(6)
Bié'n doi Fourier
ngudc SJn,n)áé
thu dUdc trUdng bié'n dói.
1.6. MÓ HINH
HOÁ
CÁC
PHÉP
TÁCH TRÜÓNG KHÁC NHAU.
II
De có thé so sánh dUdc tinmhs náng cüa các phép tách trUdng khác nhau
trong mién tan só', trong cóng trinh náy có ba quá cáu da dUdc chon. Các thóng só
cüa chúng nhU sau:
xl =
3500;yl = 3600;zl = 6000;R1 = 2500;
x2 = 2200;y2
=
2000;z2 = 300; R2 =
150;
x3 = 4000;y3 = 4000;z3
=
300;R3 =
150;
trong dó xi,
yi,
zi tUdng úng lá toa do trong tám cüa các quá cáu. Ri lá bán
kinh
cüa
các quá cáu. Mat do du lá 500
kg/m^.
Hieu úng trong
lUc
cüa
dóng
thdi ba quá cáu dUde thé hien
tren
H.l.
Tú các
só"
liéu nguyén thuy
tren
H.l, theo cóng thúc (7) vá các
lenh
vé bien
dói Fourier trong Mathematica, các phUdng pháp tách trUdng khác nhau dUdc
thUc hien. Các két quá tính
toan
dUdc trinh báy tren các hinh H.2
den
H.7
- Tié'p tuc giái tích xuó'ng 250 m (H.2).
- Tié'p tuc giái tích lén
tren den lOOOm
H.3).
Dao hám tháng dúng hang hai (H.4).
- Dao hám tháng dúng hang ba (H.5).
60
50
^
40
3
30
>^20
10
O
I
!i
I
'I
O 10 20 30 40 50 60
x.lOO
m
60;;,
50|L
?
!
i
H
30!;;
>^20j''
lOi
O'
O 1020 30 4050 60
x.lOO
m
II.1
Hiéu úng
tr9ng
h/c
cua quá
cáu
H.2. Tiep tuc giái tích xuó'ng
dUói 250ni
60¡
50
o 40
5
30
>20
10
O'
O
10 20 30 40 50 60
X.lOO
m
60 ;
50
V
40;'
H
30'
>^20
10
O
O 102030 4050 60
X.lOO
m
H.3.
Tiep tuc
giái
tich xuó'ng
dUói lOOOm
H.4, Dao hám bác hai
60;;Ji'
lí^sl'ií!';"
50
40
H 301
>'20'
10
O
V''
O 10 20 30 40 50 60
X.lOO m
100
m
>i
gQ'5l,£^'íí<-<^,-?.v'
50^jf •
30']::i:.'':'''/?''
20i
:.í'-
'^'^ ^
V^r
;•>>•->v^:
I',
I:
lOl'i'
:V
r^-^-
0 1020 30 4050 60
X.lOO m
H.5.
Dao hám hang ba H.6 Bien doi Saxov-Nigard vói các bán
kinh
2000-2500
ni
o
o
60
50
40i
30'
20;
loi
O'
O 10 20 30 40 50 60
X.lOO
m
H,7.
Bien doi Saxov-Nigarrd vói các bán
kinh
5000-7000 m )
Bien doi Saxov-Nigard vói các bán
kinh
2000-2500 m (H. 6)
- Bié'n doi Saxov-Nigard vói các bán
kinh
5000-7000 m (H. 7)
1.7.
KÉT
LUÁN
Tú hieu úng trong
lUc
tong cüa cá ba hinh cáu, sü dung phán mém
"Mathematica" [4], theo thuát
toan
tính trUdng trong mién
tan
só, các thánh
phán trUdng do các vat thé d các do sáu khác nhau dUdc phát hien.
Vói thuát
toan dá
dUdc trinh báy cúng vói táp hdp các lenh cüa
Mathematica, ta có thé trUc tiép nghién cúu dUdc trUdng tong theo các
thánh phán trUc tiép
tren
các máy tính cá nhán thóng thUdng.
Thuát
toan
cüng vói Mathematica
sé
trd giúp nhiéu cho các nhá nghién cúu
cá'u truc
trUÓng
thc ngay tai cd sd
lám
viec má khóng nhá't thict qua các
trung tám xü ly
so'lieu
cao cá'p.
TAI LIÉU
THAM KHÁO
1.
Ton Tich Ai.
Computaíional
Method (
Phuon^
phap
so ). National
Univcrsity
Publisher, Hanoi 2000.
2.
Ton Tich Ai. Applied Geophysics ( Dia vat ly tham do)
.Univcrsity
Ministry
Publisher, Hanoi 1988.
J.
Stephen Wolfram.
Matemática.
Addison-Wesley
Publishing Company, Inc.
1988.
4.
Tafeev.G.PjSokolov
K.P.
Geological interpretation of magnetic anomalies. Nedra.
Leningrad.1981.
5. Nikitin
A.A.
Síatistical Processing of Geophysical Data. Electromagnetic.
Research
Center. Moscow
1993.
Bao cao
tai hoi nghi quó'c té': International
Coference
on High
Performance Scientific Computing.
Modeling, Simulation and Optimization of Complex Processes.
March 10-2003 Hanoi, Vietnam
Mathematica for
Geophysical
Fieid Separation
Ton Tich Ai
Abstract:
The
articie
contains
different techniqíies
of geophysical data
processhig
by using software
Mathematica. It was known that geophysical data are usually
dislorted
due to the
influence
of noisc.
Predominantly, the noise can be caused by geological inhomogeneous
strwctuie,
by
errors
in measurements and
methods being used, by variations of physical
fields,
etc. As a matter of fact, geophysical data processing ¡s
ihe
combat with the noise of various natures. In geophysical data processing, the field separation acquires great
signifl
canee.
Field separation is one of the most important problems ¡n geophysical data interpretation. For potential
fields,
when there are observational data for the both the
proflle
survey and
área
survey. The field separation
becomes a process
ofestimation
of
low
frequency component, i.e., the regional anomaly, on the one hand, and
high- frequency field componcnt, i.e. the residual or local anomaly on the other hand.
Author, in frequency domain
establishes
the kernel of Saxov Nigarrd Transform, and this method is
sampled
for comparison with other methods. The
results
of sampling
ailow
us lo
use
thc
method
in
piacticc.
In this articie the field
separation
is realized in frequency domain by using Computer Algebra
system
Mathematica.
Introduction
The numerical simulation
of
various natural and tcchnological
processes
or
phenoincna
with the aid of computers has now bccome a
powcrful tool
for studying
tlicsc processes
or
phenomena.
In applled
geophysics the numerical simulation
Is
based on using different
software for processing, separating geophysical data.
In many
cases, the dctcrmining
parameters of bodies, causing geophysical anomalies is
main
process in studying practical
geophysical data.
Field separation is one of the most important problems in geophysical data
interpretation. For potential
ílelds,
when there are observational data for thc both thc
prolllc
survey and
área
survey. The field separation becomes a process
ofestimation
of low
frequency
component, i.e., the regional anomaly, on the one hand, and high- frequency field componcnt,
i.e. the residual or local anomaly on the other hand
[51.
At present, numerous processing techniques for field separation are used dcmand
various a priori Information on window size for data smoothing, on spectral composition of
regional and local anomalies for linear filtering, on the magnitude of height for the upward
continuation and so on.
Transforming Methods of potential
fields.
The main purpose of separation (transformation) of potential (gravity and magnetic)
fields is the extraction from
observed
field into components that associated with individual
geological objects locating at different depths. The
resultíng
function, depending on
transforming operations may be of the same unit of initial function or it's
dcrivatives.
Derivatives of initial function may be taken at started
level
or at new ones. The transformed
function may be have unit of product of initial function and coordinates.
Before eighty years, due to low
techníque
and poor computing tool, most problems of
field transformation were mainiy carried out in space domain. At present, these problems, in
general tendency,
will
be implemented in frequency domain with assistance of Fast Fourier or
of other software.
Transformation in Space Domain
The mathematical expression of popular field transformation in space domain can be
drawn out [2];
-Averaging:
The average of observed field is taken within the
circle
of radias R at
center
ofcircle:
,
2n ¡i
V(0,0,0) = j \\y(r,a,0)rdrda
(1)
-Analytical
continuation of thc field from O-
Icvcl
to
Z-level
is cxprcsscd by:
V(r.a,0)drda
(r'^z^)''
•Derivatives undertaken
along
x-axis
can be written
z r
r
f
r.u.ujuruu
15
2jt
wi
/ r
cK(^r,a/;;zr cosadrda
yjO.O z) =
-¡\
^-,^-jj-,
(3)
-Derivatives calculated with
respect
z-axis:
In particular way, for calculating above-mentioned transformations, the palelles
designed for that purpose are used. Furthermore these analytical exprcssions are
suitable for
coinputer
calculation.
Transformation is
carried
out in frequency domain
In
this domain
all
above-mentioned transformations can be expressed in common
formula:
For 3D problems:
VJx,y,z)= ¡¡V(4.7,.ü)K(^-x.n-y.z)d^d>i
(5)
•S
For 2D cases:
yjx,z)=
\V(4,0)K(^-x,z)d^
(6)
/.
Where
VJx,y,z),V^^(x,z)
is transformed field?
í^/í-'/.^^í^./í-^^Is
initial fields
K(^
-xjj-
y.z),
K(^
-x,z
)-
Transformation kerncls corresponding
respectively to 3D or 2D problems.
Expression (5) and (6), which are
integráis
of two functions, calleó convolution.
Mathematically, the process of signal filtering is described by such
integráis.
That is the
problems of potential field transformation can be considered as frequency filtering, in which
various transformation operations are
realized
with
different
frequency charactcristics.
According to the theory of convolution in thc frequency domain,
spectruin
SJaj)o^
lunction
y,^(x,y,z)o\'
V,Jx,z)\s
the multiplication of ones
oí V(^.¡].0),(V{^,0)) and
K(^-x,n.y,z),(K(^-x,z)),\.Q.
SJco) =
S(co)0(co)
(7)
0(co) = SJco)/S(co)
(8)
Where
S(ci))\s
spcclrum of
V(^.i],0)(V(^J})),
0(ci))\s
spcctrum
(frcqucnc\
charactcristics) of function
K(^
-
x,i],y.z)(K(^-x,z))
.
Using direct Fourier Iransform,
S,/co)
in
formula (7) is calculated and after that, by
inverse Fourier transforms, function
yjx,y,z)\s,
dctermined.
Table
1
gives
thc frequency
charactcristics for different transformations of potential data.
Table I
Transformation and its
frequency
Charactcristics
Transformation
Average:
3D Problem
2D Problem
Analytical Continuation
Upward continuation
Downward continuation
Vertical
derivatives
of n-ordcr
On the observed surface
At the height z
Horizontal derivatives of n- ordcr
On the observed surface
At the height z
Frequency Charactcristics
2J
,(cor
)/
Sin(cor )
sin( cor ) /(
cor
)
Exp(-coz)
Exp(coz)
co"
co"
exp(-ct)z)
(ico)"
(ico)"
exp(-coz)
Frequency
Charactcristics of Sacxov-Nigarrd Transformation.
In
principie, this
method
was initially proposed by Andreev and Grifin
(Fornicr
USSR)
in sixty years with aim of dctcrmining the variation componcnt (anomaly). Latcr on, this idea
was devcloped and
iniproved
by wcstern gcophysicists and the
method
was
widely
uscd. Thc
method is aimed to emphasize components of fields caused by sources locating at certain
depth.
In the space domain, this method is
described
by follovving expression:
VJO>0.0) =
V(r,)-V(rJ
(9)
r.
- r,
Where
y(r¡)
and
í^f';^
are averagcd
valúes
of thc observed field V respectively
on
circles of radius
ri
and
r2.
The
frequency
charactcristics of transformation (9) is
dctermined
as
follo\\iiig:
From thc mathematics
eourse,
wc have a formula:
•^r
S)]\^OÁU
17
, , \ =[e-'"JJcor)a}do
(10)
(r'^z'r
Therefore,
the Poisson formula in cylindricai system of coordinatc
z r
ry(r,a,0)rdrda
, •
<
*
r
M
•
r
0,-z)
=
—
\——;
^-T-;—
may be
rcwrittcn m íollownig lorm
y(ü,0,~z) =
—
^^\y(r,a,0)e-'''J,(cor)rcodrdcoda
(1 1)
27t
w
u u
Where J(,(cor) is Bessel function of first typc and zeros range.
If
we introduce the average
valué
of function
ü'í'r^on
thc circle of radius r by the
formula:
V(r) = ~^
\y(r,a,0)da
2;r
„
(12)
so formula
(II)
may be rewritten in thc form:
y(0,0,-z)=
^coe~'"'-dco ^ V(r)JJcor)rdr (13)
When z=ü
y(0,0.0)=
^codco\y(r)JJcor)rdr
(1.14)
It is analogous with formula
(13)
we can rewrite the expression of Fourier integral for
average
valué
of function
y(ri)
on circles of radius
r;:
V(r^) =
^coJ„(cor^
)dco^^V(7)J,(cor
)rdr
(15)
If we introduce:
S(co)
= ^VMJ,(cor)rdr (16)
Then the formulas
(13)
and
(15)
bccome
y(0,0,-z)=
^e-"''-S(co)codü)
(17)
y(rj= ^S(co)J„(cor)codco (18)
As we mcntioncd
before,
the Saxov-Nigarrd method in space domain (9) may
be
rewritten as
following:
rS{co}[jJcor,)-JJcorj]
yjO,0,0) = ^ —codeo (19j
18
From
(17)
and
(19),
it is evident that the kernel of transform (2) is 0(co)
=
e""''
and
the kernel of Saxov-Nigarrd transformation is:
^^^^^J>^^^W¿í^
(20)
Transforming
gravity
field
by Saxov- Nigarrd
method
in
frequency
domain
The objective of the Saxov-Nigarrd transform is studying the structurc of interested
región
at different depths by variation parameter
ri
and
r2
in the formula (20).
From Eq. (20) and (7), in practice, in order to realize thc Saxov-Nigarrd transform ¡t is
ncccssary to
apply
the following algorithm.
(1)
Construction of field or
model
data matrix n-n
dimensión
daía(n,n)
(2)
Direct
Fourier transform of da{a(n,n) for receiving matrix of Fourier Transform
S(n,
n)
(3) According to formula (20), thc kernel matrix
0(n,n)of
frequency charactcristics is
calculated, For
otlier
transformations, 0(n,n) is calculated on basics of
frequency
charactcristics from Table
1.
(4) Finding frequency matrix of transformed field
S^/n,n)by
multiplication
S(n,n)by
0(n,n).
(6)
Invcrsc Fourier transforming
S^Jn.n)
for receiving transformed field.
Modeling
different
fíclü
separatíons.
To
enable
comparison between different methods of field separation in frequency
domain, in this articie the model of thrce spheres is choiced. The parameters of these sphcres
are follovving:
X i
= 3500;y
1 =
3600;z
1
= 6000;R
i
= 2500;
x2
=
2200;y2 = 20ü0;z2
=
300; R2
=
150;
x3 = 4000;y3 = 4000;z3
=
300;R3
=
150;
Where xi, yi, zi are respectively the coordinates of centers of spheres, Ri are radii of
the spheres. The excess density of these spheres is 500
kg/m3.
'fhc
gravity clTect of these spheres is presented on fig. I
JM'om
original data in Fig. I, by formula (7) and Fourier Transform in
Mathematica.
different methods of field separation are
realized.
Results
ofthe
calculation are
presented
on
F¡g.2-Fig.7:
Dov\in\ard
continuation to 250 m (Fig. 2).
Upward continuation to
lOOOm
(Fig.
3).
Vertical derivalive
2-ordcr
(Fig. 4).
Vertical derivalive 3-order (Fig. 5).
60
50
40
H
30!
>^20
10
0^
O 10 20 30 40 50 60
X.lOO
m
60;
,i,
5o;:::'''
o40V
S30
>^20^
10'
O'
o
10 20 30 40 50 60
X.lOO
m
Fig. 1 Gravity
field
of threc spheres
Fig. 2 Downward continuation to 250 m
60!
50
.
B
'
o
i,
^30!
>^20Í
10;
0^
0
§£-•'
i:i\;-;:¡r'i
10 20 30 40 50 60
X.lOO
m
60,
,
50;:.
o4ol;
S
30;:
>^20^
10
0
0
o
102030 4050 60
X.lOO
m
Fig. 3 Upward continuation to
lOüOni
Fig. 4 Vertical dcrivative 2-ürdcr
60Í/
5o|;,
40i;:'
H
30¡.
>^20Í
lOi
O-
o
10 20 30 40 50 60
X.lOO m
60
50
40
H
30
>^20
10
O
O 1020 30 4050 60
X.lOO m
Fig. 5
Vertical
dcrivative 3-ordcr Fig. 6 Saxov-Nigarrd Transform with radii
20Ü0-
500 m
20
60
50
^
40
o
S30
-20
10
0
i,
•\ •
\
•'• •
•
-ir'
o
10 20 30 40 50 60
X.lOO
m
Fig. 7 Saxov-Nigarrd Transform with radii
5000-7000 ni
Saxov-Nigarrd Transform with radii 2000-2500 m (Fig. 6)
Saxov-Nigarrd Transform with radii 5000-7000 m (Fig. 7)
Conclusión
From total gravity effect of three different
dimensión
spheres, usíng
software
"Mathematica" [4], by discusscd algorithm oí' field separation in frequency domain, the field
components,
caused
by bodies at
different
depth are picked out.
Presented algorithm, realized on basis of Mathematica commands, allows us directly to
investígate the total field on ordinary computer.
References
6. Ton Tich A¡, Computaíional
Meííwd
( Phuong phap so
),
National Univcrsity
Publisher, Hanoi 2000.
7.
Ton Tich Ai. Applied Geophysics ( Dia val ly íham do)
.Universily
Ministry
Publisher, Hanoi 1988.
8. Stephen Wolfram.
Matemática.
Addison-Wesley Publishing Company, Inc. 1988.
9. Tafcev.G.P,Sokolov K.P. Geological interpretation of magnetic anomalies. Nedra.
Leningrad.1981.
10.
Nikitin
A.A.
Statisíical Processing of Geophysical Data. Electromagnetic.
Research Center. Moscow
1993.
CHUQNG
TRÍNH
CHAY
TREN
MATHEMATICA
Clenr|x, y,
z,
xÜ,
yO,
zÜ,
dx, dy,
deltag,
Vxz,
Vzz, k,
iho,
dataü,
datal,
oiiiega,
data2,
d:ita3,
datíi3,
tia(a4,
dat:i5|
dx=
100.;
dy=
100.;
k
=
6.667
lü'^-II;
iho
= 500;
M
=
rho(4/3)Pi
U'^3;
dcltag|x_,
y_,
K_,
xO_,
yü_, zü_| =
21
k
M
zO/((x
-
x0)^2 +
(y -
y0)^2
+
z0'^2)'^(3/2);
Vxz|x_,y^,
R_,xO_,yO_,
zO_|
= -3 k
M
zO
(x -
xO)/((x
-
x0)'^2
+ (y -
y0)^2 +
z0''2)^(5/2);
Vzz|x_,
y_,
R_,
xO_, yO_,
z0_|
=
k
M
(2z0'^2
- (x -
x0)'^2)/((x
-
xO)^2
+ (y -
y0)'^2
+
z0^2)^(5/2);
Clear|xl,
yl,
zl,
x2, y2,
z2,
RI,
R2|
xl
=3500;
y I
=3600;
zl
=4000;
RI =2500;
x2
-
2200;
y2
=
2000;
z2
= 300;
R2
= 150;
x3
=
4000;
y3
=
4000;
z3
=
300;
R3
=
I50;
rl
-2000;
r2
=
2500;
flx_,yj =
I0^5(dcltag|x,
y, RI, xl, yl, zl | +
dcltag|x,
y, R2, x2, y2, z2| +
deltag|x,
y, R3,
x3, y3,
z3|);
dat:iO
=
Tablc|f|idx,j
dy|, {i,
64}, {j,64}l;
díO
=
ListContourPIot|dataO,
ContourShading
->
False,
Contours
-> 20,
FraincLabel ->
{"x.lOO
m",
"y.lOO
m"},
ContourStyle->
RGBColorll,
O,
IIJ;
dataO
=
Fourícr|dntaO];
omcga =
Sqrt|(i/(64
dx))^2
+
G/(í»4dy))^2|;
kcrnell
=
Tablc|Exp[250
onicga|
, {i,
64}, {j,
64}t;
kernel2
=
Tablc|Exp|-IOOO
omcga]
, {i,
64},
{j,
64}|;
kcrncl3
=
Tablelomcga ,
{i,
64}, {j,
64});
kcrnel4
=
Table|omega^2,
{i,
64}, {j,
64}];
kernel
5
=
Table|(BesselJ|0, omega
rl]
-
BesscIJjO,
omega r2|)/(r2-
rl),
{i,
64},{j,64}|;
kernel
6
=
Tablc|(BcsselJ|0,
omcga SÜOO]
-
BcssclJ|0, omcga 7üüül)/(2ü00),
{i,
64},
íj,64}|;
daíal = IiiverscFouricr[dataO
kcrncll|;
data2
=
IiiverseFourier|dataÜ
Ucrncl2|;
data3
= Inverscrouricr|dataO
kernel3|;
dala4
=
InverseFouricr|dataO
keriiel4|;
data5
=
InvcrseFouricr|dataÜ
kcrncISj;
data6
=
lnvcrscFourícr|datnO
kerncl6|;
dtl
=
LislContourPlot|Abs|dataI|,
CoiitoiirSIiading
->
Falsc,
Contours->
40,
ContourStyle
->
RGBColor|l, O,
1],
FrameLabel
->
{"x.IüO
ni",
"y.lOO
m"},
DispIayFuiiction
->
Idcntity|;
dt2
=
ljsíC"ontourPI(>l|Abs|dala2|,
CoiiíourSIiading
->
l-alsc,
Contours
->
4(»,
C:ontoiirStylc
->
RGBColor|
I,
O,
11,
FrameLabel
->
{"x.IOO
m",
"y.lüü
m"},
DisplayFunction
->
Ideiitity]
dt3
=
LisfContourPlot|Abs|data3],
C'ontourShadiiig
->
Falsc, Contours
-> 40,
ContourStyle
->
RCBColorlO,
O,
Ij,
FrameLabel
-> {"x.IOO
m",
"y.IOO
m"},
DisplayFunction
->
Idcntily|;
dt4
=
ListConíourPloí|AI)s|data4|, ContourShading
->
Falsc. Contours
-> 40,
ContourStyle
->
RGBCüIor|0,
ü,
I|,
FrameLabel
->
{"x.lUU
in".
"y.IOO
m"}.
")0
DisplayFunction •>
Idenlityl;
dtS =
ListContourPlot|Abs|data5|,
ContourShading -> False, Contours -> 40,
ContourStyle -> RGBCoIor|I,
O,
l|, FrameLabel -> {"x.IOO m", "y.IOO
m"},
DisplayFunction -> Identityl;
dt6 =
ListContourPlot|Abs|data6|,
ContourShading -> Falsc, Contours -> 40,
ContourStyle->
RGBColor|I, O,
Oj,
FrameLabel->
{"x.IOO
m",
"y.IOO
m"},
DisplayFunction ->
Idcntityj;
Show|GraphicsArray|{dtO,
dtl}|, DisplayFunction ->
SDisplayFunctionj;
Show|GraphicsArray({dt2,
dt3}l,
DisplayFunction ->
SDisplayFuncíionj;
ShowIGraphicsArray|{dt4,
dt5}|, DisplayFunction ->
SDisplayFunctionj;
Show|Graph¡csArray|{dt6}j,
DisplayFunction ->
SDisplayFunctionl;
PHAN II
PHÁN CHIA
TRÜÓNG
BÁNG
PHÜONG
PHÁP CÁC YÉU TO
CHÍNH.
Muc dích cua
phUcíng
pháp phán tích yéu tó" lá nghién cúu cá'u truc bén
trong
cüa
ma trán hiep bien (covariance matrix) cho mot chu5i các só' liéu có phán
bó'
ngáu nhién. các só'
lieu
náy
thiíóng
lá các só' liéu dia vát ly phán bo'
tren
N
tuyén hoac lá só' liéu cüa N
truídng
dia vát ly phán bó' tren mót tuyén hoac lá N
thóng só'cüa vat thé (do sáu, các kích thu'óc hinh hoc.) doc theo mot tuyén.
Triíóng
hdp thú nhá't
lá trUcíng hcíp
phán chia
triídng,
con tru'dng hdp thú
hai lá
trUÓng
hdp thú hai thuoe
vé
bái
toan
tu chinh các só' liéu, con tru'dng hdp
thú ba lá
tnJdng
hdp phán tích dinh iu'dng các só' liéu día vát ly.
Trong cóng trinh náy chü yéu ta xét
den van dé
sü dung Mathematica theo
húóng
thú nhá't.
2.1.
MÓ HÍNH
TOAN
HOC CUA PHÜONG PHÁP PHÁN TÍCH YEU TO
CHÍNH. (MATHEMATICAL MODEL OF FACTOR ANALYSIS).
So'lieu
triídng
quan sát
diídc tren
N tuyén có thé
dUdc
biéu dién duói dang
ma trán hinh chü nhát:
X
A',
\^N
J
In
(2.1)
trong dó n lá só'diém
tren
mói tuyén.
23
Dé
có
dUdc
m yéu tó'
mói,
ngiídi
ta
phái nghién
cúu cá'u
truc
bén
trong
cüa
ma trán thóng
tin
B=X'X
dude
thánh
lap qua các dai
lüdng
tuán theo phán
bó'
chuán
X^
,X2, ,X^.
B
=
X'X =
¿^^2^\
2J-^2'^2
•••
¿_^^2^N
(2.2)
Các
yéu
tren
düdng chéo
cüa ma
trán
(2.2) lá các
phiídng
sai
(phüdng
sai)
cüa
các dai
liídng
ngáu nhién
X„
con các yé'u
tó' ngoái düdng chéo
lá các
hiep bié'n
cüa chúng
(
covariancies).
Ma
trán thóng
tin B có thé
áxiúc
chuyén thánh
ma
trán
tiídng
quan:
R
=
^1
f\,
/-^
2N
V'A'I
''22
(2.3)
trong
dó
rjj
-
i'j¡
lá các he
só'
tUdng
quan
gii][a
các dai
Itídng
ngáu nhién
X, vá
Xj.
Chúng
diídc
viét
lai
düói
dang
sau:
Y^(X,-X,)(X,-X^)
'•'/
=
k
= \
(2.4)
ncr,a
trong
dó
a¡
and
QJ
tiídng
úng lá các dó
léch
chuán tu'dng
úng vói các dai
lúdng
ngáu
nhién
X^
vá
Xj.
Muc dích chính
cüa
phu'dng pháp phán tích
các yéu
tó'chính
lá
chon
mot só'
bé
các dai
lüdng fj
dé
giái thích diídc
cá'u
truc
bén
trong
cüa các ma
trán
B
hoac
R
vá bié'n
doi các ma
trán
náy
thánh
các ma
trán
dUdng
chéo
vi các yéu
tó'f^,
f^,
,
f,,^
lá
các yéu
tó'dóc
láp
vói
nhau.
Mó
hinh
toan
hoc cüa
phUdng
pháp phán tích
yéu
tó'có
thé
du'dc vié't
diídi
dang
sau:
A;=IK./;+^'.>
(i=\,
2 N)
(2.5)
m
< N
trong
dó fj lá yéu
tó'
thú j vá tát cá các yéu tó
f^
dóc láp vói
nhau
vói sai só
e,
2.2.
THUÁT
TOAN
PHÁN TÍCH
CÁC YEU TO
CHÍNII.
24
Trong
thüc
té',
phiídng
pháp phán tích các yéu tó' chính diídc sü dung dé
phán chia triídng vá tích hdp quá trinh xü ly so'lieu. de chuyén qua các dai liídng
ngáu nhién mói dó'i vói mó hinh (2,5) ta có thé dúng các dai
lúdng
phu'dng sai cüa
chúng.
Yé'u tó' chính diídc xác dinh tü to hdp tuyén tính các dai liídng ngáu nhién
ban dáu X:
Yj=JLc,jX,
.j
=
\
A^
(2.7)
(=1
Trong triídng hdp náy yéu
tó'
chính dáu tién
r,
=¿a„x,
=^^
(2.8)
có
phiJdng
sai
cüc
dai trong tá't cá các to hdp (2.7) có thé có. Các
phiídng
sai c-+
üa các to hdp tuyén tính
Y^
diídc sáp xép theo thú tü giám
dan,
túc la:/*
a'(Y,)>G'(YJ> >a'(Y,)
Ngoái ra
ngiídi
ta con sü
diing sií chuán
hoá sau day
dói
vói các he só'
chuyén
a^j:
Z<=1
(2.9)
Ró ráng ráng
VÍ^J
=a'Ia,
má trong dó I lá ma trán ddn vi,
a'vá
a
Udng
úng lá các vectd háng vá cot. Néu
nhií
B lá ma trán phiídng sai cüa X (2.2) vá MX
lá ma trán ky vong
toan
hoc cüa X, thi phiídng sai cüa
Yj
có thé diídc biéu dién
nhií
sau:
DV =
M(Y^
-
MY^
/ =
M(Y^
-
MY^
)(Y^
- MY^) =
—
- - -^ ^
- (2.10)
= M(aj
X-a^ADC)'(a^X-ajMX)
=
a^Ba^
Chuán
hoá (2.9) den (2.10) ta thu du'dc ty só'sau:
OBV,
A,
=^-^
(2.11)
aja^
De thu du'dc giá tri cuc dai cüa (2.11), túc lá thu du'dc giá tri
cúc
dai cüa
phiídng sai
Yj
ta phái tính dao hám cüa (2.11) theo a vá cho dao hám náy báng
khóng, cu thé lá:
25