Professional Documents
Culture Documents
Oktem Levy Type Fourier 2007
Oktem Levy Type Fourier 2007
Oktem Levy Type Fourier 2007
www.elsevier.com/locate/compstruct
b,*
Department of Naval Architecture and Marine Engineering, Yildiz Technical University, 34349 Besiktas-Istanbul, Turkey
b
Department of Materials Science and Engineering, University of Utah, 1225 S. Central Campus Drive,
Room 304, Salt Lake City, UT 84112-0560, United States
Available online 2 August 2006
Abstract
A hitherto unavailable Levy type analytical solution to the problem of deformation of a nite-dimensional general cross-ply thick
doubly curved panel of rectangular plan-form, modeled using a higher order shear deformation theory (HSDT), is presented. A solution
methodology, based on a boundary-discontinuous generalized double Fourier series approach is used to solve a system of ve highly
coupled linear partial dierential equations, generated by the HSDT-based general cross-ply shell analysis, with the SS2-type simply supported boundary condition prescribed on two opposite edges, while the remaining two edges are subjected to the SS3-type constraint.
The numerical accuracy of the solution is ascertained by studying the convergence characteristics of deections and moments of a moderately thick cross-ply spherical panel. Hitherto unavailable important numerical results presented include sensitivity of the predicted
response quantities of interest to lamination, lamina material property, and thickness and curvature eects, as well as their interactions.
2006 Elsevier Ltd. All rights reserved.
Keywords: Thick laminate; Higher order shear deformation theory (HSDT); Doubly curved panel; Spherical shell; Cylindrical shell; Cross-ply; Analytical
solution; Boundary discontinuous double Fourier series; Boundary constraints
1. Introduction
Recent years have witnessed an increasing use of
advanced composite materials (e.g., graphite/epoxy, boron/
epoxy, Kevlar/epoxy, graphite/PEEK, etc.), which are
replacing metallic alloys in the fabrication of at/curved
panels because of many benecial properties, such as higher
strength-to-weight and stiness-to-weight ratios (resulting
in fuel economy), longer inplane fatigue (including sonic
fatigue) life and stealth characteristics (of military aircraft,
e.g., stealth ghter, F-117A Nighthawk and B-2 bomber),
enhanced corrosion resistance, and so on. Since the matrix
material is of relatively low shearing stiness as compared
to the bers, polymeric composite shell type structures are
highly prone to transverse shear related fatigue failures. A
reliable prediction of the response of these laminated shells
or doubly curved panels must account for transverse shear
deformation. Additionally, a solution to the problem of
*
Corresponding author. Tel.: +1 801 581 6282; fax: +1 801 581 4816.
E-mail address: r.chaudhuri@m.cc.utah.edu (R.A. Chaudhuri).
0263-8223/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2006.05.020
476
1
g
e4 n1 ; n2 ; n3
u3;2 2 u2 u2;3 ;
n3
R
2
1 R2 g2
1
g1
u3;1 u1 u1;3 ;
e5 n1 ; n2 ; n3
R1
1 Rn31 g1
1
1
u2;1 g1;2 u1
e6 n1 ; n2 ; n3
g2
1 Rn31 g1
1
1
u1;2 g1;2 u1 ;
g2
1 Rn32 g2
1c
1d
1e
1f
u2
f
4f3
1
1
u2 f/2 2 /2 u3;2 ;
R2
g2
3h
u3 u3 ;
2b
2c
6a
f2 j21 ;
3a
e2 e02 fj02
f2 j22 ;
3b
e1
e01
fj01
e4 e04 f2 j14 ;
3c
e5 e05 f2 j15 ;
3d
3e
in which
e01 u1;1
u3
;
R1
4a
j01 /1;1 ;
4
j21 2 /1;1 u3;11 ;
3h
u3
0
e2 u2;2 ;
R2
j02 /2;2 ;
4
j22 2 /2;2 u3;22 ;
3h
e04 u3;2 /2 ;
4
j14 2 /2 u3;2 ;
h
e05 u3;1 /1 ;
4
j15 2 /1 u3;1 ;
h
e06 u2;1 u1;2 ;
4b
j06
4l
/2;1 /1;2 ;
4
j26 2 /2;1 /1;2 2u3;12 :
3h
4c
Q2
A4j e0j
K1
D5j e0j
D4j j1j ;
F 5j j1j
i; j 1; 2; 6
6b
6c
6d
6e
j 4; 5
6f
6g
4d
477
Eij ; F ij ; H ij
4e
N
X
k1
fk1
fk
Qij f3 ; f4 ; f6 df
7b
fk1
k
4f
4g
4h
4i
4j
4k
4m
8a
a5 u3;11 a8 u3;22 ;
N 6 A66 u2;1 A66 u1;2 a9 /2;1 a9 /1;2 a10 u3;12 ;
8b
8c
b4 u3;11 b5 u3;22 ;
M 2 B12 u1;1 b6 u3 B22 u2;2 b3 /1;1 b7 /2;2
8d
N 1;1 N 6;2 0;
5a
N 6;1 N 2;2 0;
5b
4
4
Q1;1 Q2;2 2 K 1;1 K 2;2 2 P 1;11 P 2;22 2P 6;12
h
3h
N1 N2
q;
5c
R1 R 2
4
4
5d
M 1;1 M 6;2 Q1 2 K 1 2 P 1;1 P 6;2 0;
h
3h
4
4
M 6;1 M 2;2 Q2 2 K 2 2 P 6;1 P 2;2 0;
5e
h
3h
b5 u3;11 b8 u3;22 ;
M 6 B66 u2;1 B66 u1;2 b9 /2;1 b9 /1;2 b10 u3;12 ;
8e
8f
8g
8h
8i
Q1
Q2
K1
K2
d 2 /1 d 2 u3;1 ;
d 1 /2 d 1 u3;2 ;
d 4 /1 d 4 u3;1 ;
d 3 /2 d 3 u3;2 :
8j
8k
8l
8m
478
The constants ai, bi, di, referred to in Eqs. (8) are given in
Appendix A.
Substitution of Eqs. (8) in equilibrium equations, given
by Eqs. (5), supplies the following ve highly coupled
fourth-order governing partial dierential equations:
9a
9b
0 < x1 < a; 0 6 x2 6 b;
1 X
1
X
u2
V mn sinax1 cosbx2 ;
0 < x1 < a; 0 6 x2 6 b;
1 X
1
X
u3
W mn sinax1 sinbx2 ;
12a
12b
m1 n1
0 6 x1 6 a; 0 6 x2 6 b;
1 X
1
X
X mn cosax1 sinbx2 ;
/1
0 6 x1 6 a; 0 6 x2 6 b;
9c
12c
/2
1
X
1
X
12d
Y mn sinax1 cosbx2 ;
m1 n0
9d
U mn cosax1 sinbx2 ;
m0 n1
m0 n1
1 X
1
X
m1 n0
u1
9e
10a
M 1 0; x2 M 1 a; x2 0;
10b
/2 0; x2 /2 a; x2 0;
10c
u1 0; x2 u1 a; x2 0;
10d
N 6 0; x2 N 6 a; x2 0;
10e
P 1 0; x2 P 1 a; x2 0:
10f
11a
M 2 x1 ; 0 M 2 x1 ; b 0;
11b
/1 x1 ; 0 /1 x1 ; b 0;
11c
N 2 x1 ; 0 N 2 x1 ; b 0;
11d
u1 x1 ; 0 u1 x1 ; b 0;
11e
P 2 x1 ; 0 P 2 x1 ; b 0:
11f
3. Method of solution
The displacement functions (particular solution) are
assumed to be in the form [18,20,24]:
0 6 x1 6 a; 0 6 x2 6 b;
12e
where
a
mp
;
a
np
:
b
13
u1;1
1 X
1
X
aU mn sinax1 sinbx2 :
14b
m1 n1
1
1 X
1
X
X
2
1
a U mn cm
an wm bn
an sinbx2
2
m1
m1 n1
cosax1 sinbx2 :
15
1
X
cn cosbx2
n1
wm d0 cosax1
1
2
1 X
1
X
aV m0 cmc0
aV mn cmcn
16
0; 1; m odd;
17
1; 0; m even;
n ,
bn , cn and
and where the boundary Fourier coecients a
dn , are as dened in Appendix B. The remaining particular
solution functions or their partial derivatives do not have
any boundary discontinuities, and therefore, can be dierentiated term by term. This step generates additional
4n + 2 unknowns.
Introduction of the displacement functions and their
appropriate derivatives into the governing partial dierential equations will supply 5mn + 2m + 2n equations as
given below:
1 X
1
X
cosax1 sinbx2
m1 n1
a2 A11 b2 A66 U mn
3
m1 n1
m1 n1
18b
a6 b a8 b3 f8 a2 b V mn
f6 a2 f15 f7 b2 f11 a4 f12 a2 b2 f14 b4 W mn
f4 a f9 a3 f10 ab2 X mn f5 b b13 b3 f10 a2 b Y mn
a4 acm
an w m
bn
1 X
1
X
qmn sinax1 sinbx2 ;
18c
m1 n1
m1 n1
a9 a2 a7 b2 V mn e12 b e7 a2 b e10 b3 W mn
e4 abX mn e5 a2 e11 e9 b2 Y mn
aa9 cmcn wm dn 0;
sinbx2
18e
A66 2
A11
an 0;
b U 0n a9 b2 X 0n
2
2
19a
19b
ha
i
a2
9
sinbx2 b2 U 0n e8 e5 b2 X 0n an 0;
19c
2
2
n1
1
ha
i
X
a9
9
sinax1 a2 V m0 e11 e5 a2 Y m0 acmc0 wm d0 0:
2
2
m1
1
X
19d
abf1 V mn a2 a f3 ab2 a4 a W mn a9 b2 a2 a X mn
abf2 Y mn A11 cm
18a
an w m
bn 0;
1 X
1
X
sinax1 cosbx2 fabf1 U mn
A66 a2 A22 b2 V mn a6 b f3 a2 b a8 b3 W mn
f2 abX mn a9 a2 a7 b2 Y mn
aA66 cmcn wm dn 0;
1 X
1
X
sinax1 sinbx2 a1 a a4 a3 f8 ab2 U mn
18d
A66 2
A66
2
a V m0 a9 a Y m0
acmc0 wm d 0 0;
sinax1
2
2
m1
m1
in which
cm ; wm
abe1 V mn e6 a3 e7 ab2 e2 a W mn
e8 e3 a2 e5 b2 X mn e4 abY mn
a2 cm an wm bn 0;
1 X
1
X
sinax1 cosbx2 fabe1 U mn
n1
1
X
m1 n1
cosax1 sinbx2 a2 a2 b2 a9 U mn
m1 n1
1
X
1
X
1 X
1
X
479
U 0n
1
X
cm U mn 0:
20b
m1
bX mn a9 H
abW mn a10 H 0;
21a
m1
X
A66
m1
m1
aV m0 cmc0 wm d0 H aY m0 a9 H
A66
c0 0;
4
21b
480
START
Calculate
Deflections
Moments
etc.
-Number of
plies
-Thickness
of plies
-Lamination
Angles, etc.
-Calculate Stiffness
(A, B, D, E, F, H
Matrices)
-Calculate other
constants
YES
STOP
Number of
Terms for
Calculation
( an , bn , etc.)
Fig. 2. Flow chart for numerical solution.
m
fA66 aV mn bU mn aY mn a9 bX mn a9
m1;3;5...
abW mn a10 A66 dn wm 0;
22a
1
X
fA66 aV mn bU mn aY mn a9 bX mn a9 abW mn a10
m2;4;6...
A66
A66cn gcm
cn bU 0n bX 0n a9 0;
22b
2
1
X
A66
aV m0 cmc0 wm d0 aY m0 a9 wm 0;
2
m1;3;5...
1
X
A66
aV m0 cmc0 wm d0 aY m0 a9 wm
2
m2;4;6...
A66
c0 0:
4
22c
22d
103 E2 h3
u3 ;
q0 a4
M 1
103
M 1;
q0 a2
in which a is assumed equal to 812.8 mm (32 in.) and q0 denotes the uniformly distributed transverse load and equals
to 689.5 kPa (100 psi). u3 and M 1 are computed at the center of the panel with the exceptions of Figs. 19 and 20.
Before presenting numerical results for spherical and
cylindrical panels, those pertaining to at (R1 = 1 and
R2 = 1) symmetric and antisymmetric cross-ply plates
with the same boundary conditions have been reproduced
rst [24]. Fig. 3 displays the convergence (with m = n) of
normalized transverse displacement (deection), u3 and
moment, M 1 , of a moderately thick (a/h = 10) and moderately deep (R1/a = R2/b = 10) antisymmetric cross-ply [0/
90] spherical (R1 = R2) panel of square (a = b) plan-form,
computed using the present HSDT with the material type I.
Rapid and more or less monotonic convergence is observed
for the normalized central deection, u3 . Although the convergence plot of the central moment, M 1 , exhibits an initially oscillatory behavior, the oscillations die down very
rapidly, rendering the convergence plot practically monotonic for m, n P 10 (Fig. 3). These converged results are
in full agreement with their FSDT-based counterparts.
Material Type I 0/90 Lamination Spherical Panel
a/h=10; R/a=R/b=10
180
160
140
120
100
u3*
M*1
80
60
40
20
0
0
10
20
m=n
30
40
481
Table 1
Comparison of u3 , based on FSDT and HSDT, of symmetric and
antisymmetric cross-ply moderately deep (R/a = 10) spherical panels for
various a/h ratios and material types I and II
a/h
%Error HSDTFSDT 100
HSDT
[0/90]
3
5
10
20
[0/90/0]
[0/90/90/0]
II
II
II
3.740
2.650
1.020
0.200
4.710
2.790
0.930
0.220
18.050
17.660
10.550
4.360
11.950
9.610
4.440
1.550
20.510
18.810
11.430
4.760
12.880
9.160
3.890
1.290
Table 2
Comparison of M 1 , based on FSDT and HSDT, of symmetric and
antisymmetric cross-ply moderately deep (R/a = 10) spherical panels for
various a/h ratios and material types I and II
a/h
%Error HSDTFSDT 100
HSDT
[0/90]
3
5
10
20
[0/90/0]
[0/90/90/0]
II
II
II
6.940
1.930
0.210
0.200
3.230
0.810
0.060
0.100
8.710
6.230
2.800
0.410
2.710
2.350
1.090
0.190
11.890
9.680
4.600
1.050
0.300
1.510
0.980
0.220
Tables 1 and 2 present the variation of central deection, u3 , and moment, M 1 , of antisymmetric [0/90] and
symmetric ([0/90/0] as well as [0/90/90/0]) cross-ply
moderately deep (R/a = 10) spherical panels computed on
the basis of FSDT, and HSDT shell theories, with respect
to a/h ratio for the aforementioned two sets of material
properties, respectively. The shear correction factors,
K 21 K 21 5=6 are assumed in the present FSDT computations. The percentage of error is dened by
HSDT FSDT
100:
%Error
HSDT
Relative inaccuracy of the FSDT with respect to the HSDT
is self-evident in the case of thick (a/h < 10) symmetric
cross-ply panels, while the former may, in general, be considered acceptable for thick antisymmetric [0/90] panels
(Table 1). The FSDT may also be considered acceptable
for moderately thick (10 6 a/h 6 20) antisymmetric
[0/90] and intermediate modulus (material type II) symmetric laminates (Table 1). Table 1 further shows that
the FSDT based u3 values for high modulus (material I)
symmetric laminates are in substantial error (more than
10%) with respect to their HSDT counterparts even for
a/h = 10. The dierence between two theories increases
with the increase of a/h ratio (for thick and very thick regime) and as well as the number of layers increase as seen
in Tables 1 and 2. The reason lies in the fact that the eect
of interlaminar shear (unmatched properties through the
thickness) increases with the increase of the number of layers. It is further noteworthy from Table 1 that a far more
pronounced thickness shear eect is observed in the computed central deection of symmetric ([0/90/0] and
482
to their HSDT counterparts for materials I and II, respectively. Additional results are shown in Table 1. There is
reason to believe [8,25] that eect of thickness is compensated, to a certain extent, by the bendingstretching coupling eect, a characteristic of antisymmetric laminates.
Same conclusion can be inferred for the dierence between
the two theories with regard to the computed central moment, M 1 , as shown in Table 2. These results conrm the
earlier assertion that eect of thickness is compensated,
to a certain extent, by the bendingstretching coupling effect, a characteristic of antisymmetric laminates [8,25].
Figs. 4 and 5 present the variations of central deection,
u3 , of [0/90] spherical panels, with a/h and R/a ratios,
Spherical Panel Material Type I a/h=10
30
R/a=20
R/a=60
R/a=100
Plate
25
16
u*3
u3*
20
18
0/90
0/90/0
0/90/90/0
14
15
12
10
10
8
5
0
10
20
30
40
50
a/h
Fig. 4. Variation of normalized central deection, u3 , with a/h ratio, of an
antisymmetric cross-ply [0/90] spherical panel for dierent R/a ratios.
20
40
60
80
100
R/a
Fig. 6. Variation of normalized central deection, u3 of a spherical panel
with R/a ratio for dierent laminations.
Material Type I
0/90 Lamination Cylindrical Panel (R1=)
25
25
u3*
20
15
10
a/h=5
a/h=20
a/h=50
a/h=100
20
u*3
a/h=5
a/h=20
a/h=40
a/h=100
15
10
0
0
0
20
40
60
R/a
80
100
u3 ,
20
40
60
R2 / a
80
100
with the surface-parallel boundary constraint is particularly visible in these gures. While the cylindrical shell with
R2 ! 1 exhibits a response characteristic similar to its
spherical counterpart (Fig. 10), the other (Fig. 9) shows a
dierent behavior. Moreover, for the cylindrical shell with
R1 ! 1 type, membrane action has no eect in the thick
and moderately thick regimes as shown in Fig. 7.
Plots of central moment, M 1 , of spherical panels with
respect to R/a and a/h ratios for dierent types of laminations, are displayed in Figs. 11 and 12, respectively. The
normalized central moment is virtually independent of R/
a, for R/a > 20 in case of both symmetric and antisymmetric spherical panels. However, the normalized central
30
25
25
15
10
20
u3*
u*3
20
a/h=5
a/h=20
a/h=50
a/h=100
483
R/a=10
R/a=50
R/a=100
15
10
0
0
20
40
60
80
100
10
20
R1 / a
30
40
50
a/h
Fig. 10. Variation of normalized central deection, u3 , with a/h ratio, of
an antisymmetric cross-ply [0/90] cylindrical panel for dierent R/a
ratios.
27
140
23
130
R/a=10
R/a=50
R/a=100
u3*
21
19
M 1*
25
0/90
0/90/0
0/90/90/0
120
110
17
100
15
90
13
0
10
20
30
40
50
a/h
Fig. 9. Variation of normalized central deection, u3 , with a/h ratio, of an
antisymmetric cross-ply [0/90] cylindrical panel for dierent R/a ratios.
20
40
60
R/a
80
100
484
140
140
120
120
R/a=10
R/a=50
Plate
100
0/90
0/90/0
0/90/90/0
80
M *1
M*1
100
80
60
60
40
40
10
20
30
40
20
30
40
50
a/h
a/h
M 1 ,
10
50
120
100
R/a=10
R/a=50
Plate
M*1
80
60
40
160
10
20
30
40
50
a/h
140
M 1*
120
R/a=10
R/a=50
Plate
100
80
60
40
0
10
20
30
40
50
a/h
between the spherical and R2 ! 1 type cylindrical geometries being negligible, which is, as expected, especially true
in the atter regime (R/a > 40). Fig. 18 presents the variation of normalized central deections, u3 , of moderately
deep (R/a = 10) antisymmetric [0/90] cross-ply (one
spherically and two cylindrically) curved panels with
respect to a/h. Again, the response of the [0/90] spherical
panel is almost identical to its cylindrical counterpart with
R2 ! 1 in the entire range of the a/h ratio.
Variations
response
quantities
of
of the normalized
3
interest, u1 u1 10q Ea24h u1 , u3 , /1 /1 10q Ea23h /1 , and
0
485
30
19
25
18
u*3
Sphere
R1 = Inf
R2 = Inf
16
u3*
20
17
15
15
10
14
13
Sphere
R1 = Inf
R2 = Inf
0
0
12
0
20
40
60
80
10
20
30
40
50
a/h
100
R/a
u3 ,
Fig. 18. Variation of normalized central deection, u3 , with a/h ratio, of
moderately deep [0/90] spherical and cylindrical panels.
5
4
10.8
u1 =5*u1*
u3 =u3* /10
M1 =M1* /30
1*
3
10.6
2
1
10.4
u3*
Sphere
R1 = Inf
R2 = Inf
10.2
0
-1 0
0.2
0.4
0.6
0.8
-2
10
-3
9.8
-4
-5
9.6
0
20
40
60
80
100
x1 / a
R/a
Fig. 17. Variation of normalized central deections, u3 , with R/a ratio, of
moderately thick [0/90/0] spherical and cylindrical panels.
486
u1 =20*u1*
u3 =u3*/10
M1 =M1* /40
1*
2
1.5
1
0.5
0
-0.5
0.2
0.4
0.6
0.8
-1
x1 / a
Fig. 20. Variations of displacements, rotation, and moment along the
center line, x2 = b/2, of a thick (a/h = 5) very shallow (R/a = 50) [0/90/
0] spherical panel.
u3*
35
0/90
0/90/0
0/90/90/0
30
25
20
0
10
15
20
25
E1 / E2
Fig. 21. Variation of normalized central deection, u3 , with E1/E2 ratio, of
a thick (a/h = 5) very shallow (R/a = 50) spherical panel for dierent
laminations.
general cross-ply thick doubly curved panel of rectangular plan-form is presented. A solution methodology,
based on a boundary-discontinuous generalized double
Fourier series approach that assures well-posedness of
the Fourier formulation and existence of the Fourier series solution, is used to solve a system of ve highly coupled linear partial dierential equations, with the SS2type simply supported boundary condition. Numerical
results presented here on cross-ply laminates demonstrate
fast convergence, and comparison with the available
FSDT-based analytical solution testies to the accuracy
and eciency of the method presented. The key conclusions that emerge from the numerical results can be summarized as follows:
(i) The CLT and the FSDT (with appropriate shear correction factor incorporated) under-predict the computed normalized deections as compared to their
HSDT counterparts in the thicker panel regime (a/
h 6 10), while the opposite is true in the case of computed normalized moments.
(ii) The FSDT may, in general, be considered acceptable
for moderately thick unsymmetric and intermediate
modulus symmetric cross-ply panels, 10 6 a/h 6 20
and beyond. The FSDT based u3 values for high
modulus symmetric cross-ply panels are, in contrast,
in substantial error (more than 10%) with respect to
their HSDT counterparts for a/h = 10.
(iii) Bendingstretching coupling prevalent in a [0/90]
type laminate has a softening eect of the beam-column type, which consequently increases the normalized central deection.
(iv) The eect of the transverse shear deformation is compensated to a certain extent by the bendingstretching
coupling eect a characteristic of unsymmetric
laminates.
(v) The bendingstretching type coupling has a highly
pronounced interaction with the type of surfaceparallel boundary restraint, imposed by the SS2
at two opposite edges. For example, the interaction
of the membrane action due to the cylindrical curvature with the surface-parallel boundary constraint
is quite noticeable. While the cylindrical shell with
R2 ! 1 exhibits a response characteristic similar
to its spherical counterpart, the other (R1 ! 1)
shows a dierent behavior. Moreover, for the cylindrical shell with R1 ! 1 type, membrane action
has no eect in the thick and moderately thick
regimes.
(vi) The radius-to-length ratio R/a has a pronounced
eect on the response of curved cross-ply panels. This
eect is progressively more pronounced for R/a < 40
for a given a/h.
(vii) The membrane action due to the eect of curvature
has a complex interaction with the bendingstretching type coupling eect, caused by the asymmetry
of lamination.
; a2 B11 2 ; a3 B12 2 ;
R1
R2
3h
3h
4E11
4E12
A12 A22
a4
; a5
; a6
;
R1
R2
3h2
3h2
4E22
4E22
4E66
a7 B22 2 ; a8
; a9 B66 2 ;
3h
3h2
3h
8E66
;
A1a
a10
3h2
B11 B12
4F 11
; b2 D11 2 ;
b1
R1
R2
3h
4F 12
4F 11
4F 12
b3 D12 2 ; b4
; b5
;
2
3h
3h
3h2
B12 B22
4F 22
4F 22
; b7 D22 2 ; b8
;
b6
R1
R2
3h
3h2
4F 66
8F 66
E11 E12
b9 D66 2 ; b10
; b11
;
2
R1
R2
3h
3h
4H 11
4H 12
4H 11
b12 F 11
; b13 F 12
; b14
;
2
2
3h
3h
3h2
4H 12
4H 22
4H 22
b15
; b16 F 22
; b17
;
2
2
3h
3h
3h2
4H 66
8H 66
E12 E22
; b19
; b20
; A1b
b18 F 66
2
2
R1
R2
3h
3h
4D44
4D55
4F 44
d 1 A44 2 ; d 2 A55 2 ; d 3 D44 2 ;
h
h
h
4F 55
A1c
d 4 D55 2 ;
h
4E12 4E66
e1 B12 B66 2 2 ;
3h
3h
4d 4 4b11
4b12
e 2 b1 d 2 2 2 ; e 3 b2 2 ;
h
3h
3h
4b13 4b18
4b18
e 4 b3 b9 2 2 ; e 5 b9 2 ;
3h
3h
3h
4b14
4b15 4b19
e6 b4 2 ; e7 b5 b10 2 2 ;
3h
3h
3h
4d 4
4b16
4b17
e8 d 2 2 ; e9 b7 2 ; e10 b8 2 ;
h
3h
3h
4d 3
4d 3 4b20
A1d
e11 d 1 2 ; e12 b6 d 1 2 2 ;
h
h
3h
f1 A12 A66 ; f 2 a3 a9 ; f 3 a10 a5 ;
4d 4 a2 a3
f4 d 2 2 ;
R1 R2
h
4d 3 a3 a7
f5 d 1 2 ;
R1 R2
h
4d 4 4b11 a4 a5
f6 d 2 2 2 ;
R1 R2
h
3h
4d 3 4b20 a5 a8
f7 d 1 2 2 ;
R1 R2
h
3h
4E12 8E66
4b12
4b13 8b18
f8
2 ; f 9 2 ; f 10 2 2 ;
3h2
3h
3h
3h
3h
4b14
8b15 8b19
4b16
f11 2 ; f 12 2 2 ; f 13 2 ;
3h
3h
3h
3h
4b17
a1 a6
A1e
f14 2 ; f 15 :
R1 R2
3h
a1
487
488