Professional Documents
Culture Documents
iri-vi«*
^t)-<\\0^-°\^V3- -7'-
/7 /^^ ^v^^
AD
TECHNICAL
LIBRARY
AMMRC TR 81-29
June 1981
Prepared for
DISPOSITION INSTRUCTIONS
./\MMRC TR 81-29
4. TITLE (and Subtitle) 5. TYPE OF REPORT 4 PERIOD COVERED
Interim Report -
MELT PURIFICATION VIA FILTRATION 1 Jan 81 - 1 Apr 81
6. PERFORMING ORG. REPORT NUMBER
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT, PROJECT, TASK
AREA 4 WORK UNIT NUMBERS
College of Engineering ,
Drexel University D/A Project: 1L162105AH84
Philadelphia, Pennsylvania 19104 AMCMS Code: 612105.H840011
11. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE
Unclassified
t5«. DECLASSIFI CATION/DOWN GRADING
SCHEDULE
17. DISTRIBUTION STATEMENT (ol the abstract entered in Block 20. il dlllerent Irom Report)
19. KEY WORDS CConf/nue on reverse side it r^ecessary and identity by block number)
Filtration Steel
Liquid metals Superalloys
Aluminum Purification
20. ABSTRACT CConHnue on reverse side il necessary and Identify by block number)
Background on the mechanisms and kinetics of filtration is discussed and the
Drexel mathematical model is further elaborated and updated to encompass filtra-
tion at the higher temperatures encountered in steels and superalloys. A novel
mathematical analysis of inclusion trajectory is also presented.
The aluininum-TiB2 study has been extended to evaluate a ceramic foam filter.
A low temperature model study of liquid inclusion filtration is described and
discussed.
Reference is made to ongoing work in superalloy and steel filtration,
DD 1 JAN 73 1473 EDITION OF 1 NOV 65 IS OBSOLETE
UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE (HTien Data Entered)
UNCLASSIFIED
SieuWITV CLAMIFICATIOM OF THII FAOtfWhw D*f Kxltnd)
Block No. 20
UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PACiE.(i>^»n Dmim Eitlcrtd)
•" ri^'',^\i-- '•
ABSTRACT
presented.
It summarizes the progress made during the period July 1st 1980 to
INTRODUCTION
2. BACKGROUND
4. PROGRESS TO DATE
6. FUTURE WORK
6.1 Aluminum filtration
6.2 Steel filtration
6.3 Superalloy filtration
6.4 Low temperature modeling
1. INTRODUCTION
the fluidity of the melt, cause internal porosity in castings, reduce mechanical
strength, ductility and fatigue resistance of the product and also result in poor
required for gteel and superalloys has not been available; however, advances in
ceramics, in combination with the need for greater product confidence, has
interest is largely confined to superalloys and high cost special steels, e.g.,
razor blade steel; however, the prospect of developing a low cost filtration
process for normal high tonnage carbon steels has certain technological and
spout, which conducts metal from the tundish to the mold, progressively plugs,
not only feasible, but will also be effective due to the rapidity of the kinetics.
On the other hand, the filtration of molten aluminum and its alloys has
been a commercial reality for many years and has traditionally been associated
with products such as aluminum foil. More recently, however, the increasing
use of aluminum in thin walled beverage cans, computer memory discs, demanding
(ii) Fluxes and salts suspended in the melt as the result of a prior
metal treatment process.
(iv) Oxide of the melt which is both suspended on top of the melt and is entrapped
within it due to tuburlence, e.g., A1„0_ in aluminum melts.
It has been found from industrial observation that the majority of the above
particles which are deleterious to product quality lie in the range of 1-30
were made by Brondyke and Hess of Alcoa and Brant, Bone and Emley of
British Aluminum.
Since then a number of techniques have been developed and are now well
have wide scale industrial application; the deep bed filter and the ceramic foam
filter.
(usually tabular alumina) through which the molten metal flows. The inclusions
deposit onto the grains of the filter medium due to diffusion, direct interception,
that surface forces are responsible for the retention of inclusions smaller than 30pm.
- 2 -
The primary mode of transport of inclusions smaller than SOym to the grain
surface of the filter medium is due to flow dynamics, while surface forces are
responsible for the retention of the inclusion in the case of lower temperature
systems, e.g. aluminum, while in the case of high temperature systems, e.g.
steels, the inclusion sinters to the filter medium . In a depth filter the
inclusions are dispersed through part or all of its volume (depth). It thus has
the advantage of having a large surface area for entrapment and can trap particles
much smaller than the pores of the filter bed. Industrially, therefore, a deep
bed filter has an extended life and is ideally suited to continuous casting
operations. At the end of its life, however, it does have the disadvantage of
The ceramic foam filter on the other hand consists of a sheet of open pore
ceramic foam approximately 5cm (2 inch) thick. Typically it would have 85%
volume void fraction, an average pore size of 1mm, with approximately 30 pores
foam with a ceramic slurry. The excess slurry is squeezed out so that only the
filaments of the foam remain coated. The resulting body is then dried and fired
ceramic replica of the original organic foam. A ceramic foam filter can be
installed prior to the casting station with minimal capital expenditure and merely
only has a limited life, the exact value of which depends on the filter area and
melt cleanliness: inclusions in the melt rapidly forming a filter cake on the
surface which increases in depth and progressively restricts flow. The filter
is thus good only for the single drop applications; however the foam slab is
relatively inexpensive, quick to change and does not require constant heating.
- 3 -
The goals of the current phase of this work may therefore be itemized
as follows:
2. BACKGROUND
the filtration of aluminum was the least difficult of the three proposed alloy
systems, hence it was the first task to be addressed. A large number of depth
filtration trials were made using the aluminum-titanium diboride system. Using
the experience gained from this work a steel filtration apparatus was designed,
A theoretical model was proposed and compared with the results from the
first phase of this investigation. Good correlation was noted and conclusions
melt flow rate. This work was extended to encompass variations in deep bed
and capture kinetics it was possible to begin establishing the criteria for
optimization.
- 4 -
To further characterize the above parameters, a low temperature model
which reinforced the conclusions of the aluminum/TiB„ system. All this initial
work has been fully detailed in the preceeding reports (A^MRC TR-80-160 and
TR-80-49).
Since this initial work, the aluminum/titanium diboride study has been
extended to evaluate an open pore ceramic filter and a system for introducing
counter-current gas flow (an industrial practice) into the deep bed filters,
flow field has been performed (using the profile of a tundish nozzle as a
reference).
the construction stage. The model system currently under investigation consists
packing.
The combined results of the above work have allowed the kinetic model to
be further refined and updated, taking into account the special conditions
melt systems, e.g. molten aluminum, where the retention of inclusions smaller
- 5 -
than SOy is by quasi-permanent surface associated effects. This model agreed
revised so that it will effectively describe this high temperature case. The
model is, at present, still limited to the short term behavior of the filter,
however, remain largely the same throughout the majority of a filter's useful
When a melt containing inclusions flows past the grains of a filter medium,
the inclusions should deposit on the surface of the grain if the filter is to
be effective. The primary mode of transport to the grain surface for inclusions
smaller than 30y is flow dynamics, while inclusions greater than 30\i have been
mechanisms as below.
90
= KC (1)
8t
Z
where
t is time
bulk liquid.
6 -
The above expression represents the rate of change of entrapped particles
K = K f (a) (2)
o
where K is a constant whose value depends on the properties of the fluid, the
size and surface morphology of both the filter medium and the inclusion. In the
aluminum-titanium diboride studies and in the low temperature models the function,
melt flow rates and the shape and size of the inclusions. The functional nature
K = K o II-
I —
am ' (3)
the inclusions sinter readily with the surface of the filter media and act as a
electrically conductive.
C.
exp ■^ e^ (^ '(9 - n) (4)
C m
c. re: "
exp((|)n) + exp -i e ^ (6 - n) - 1
'm
- 7 -
where the subscript i refers to the inlet condition and the dimensionless
parameters are:
K L
is = —— , dimensionless kinetic constant (7)
m
t function of time and distance along the filter. For a given value of 9, Eq. (4)
concentration, then n will be set to unity in Eq. (4) and the resulting
depth filter is the kinetic parameter K, as given by Eq. (3). However, the
bed retention a la (see Eq. (3)). The coefficient K is a function of the melt
m o
velocity through the filter, U , the inclusion size, d., the filter grain size,
m 1
d , and the melt system. The coefficient K can be experimentally determined
g o
by short time experiments, whereas determination of a would require experiments
of long duration.
maximum inclusion retention capacity with extended use. On the other hand.
for a filter having a very large capacity for inclusions, progressive blockage
of melt flow will occur with extended use. To determine the inclusion retention
K has been experimentally determined and its dependence on melt velocity has
system (see AMMRC TR-80-160 and TR-80-49) and is currently being extended to
steel filtration and a low temperature model with liquid inclusions. During the
Initial stages of filter use when the filter bed has been exposed to only a
few residence times, the inclusion retention capacity value of the filter is
K Z
"1^ = exp ( - ((.n ) = exp ( - -2- ) (9)
i m
At the exit end of the filter, the dimensionless distance, n = 1 and the
C K L ,
^ = exp ( - ^ ^ (10)
i m
data from many experiments can be used collectively to determine the coefficient
- 9 -
K . P.earrangement of Equation (10) gives
0
C.
In ( -T^ ) (11)
U K L ' C
mo o
1 1 , . ^O ^ + . _ V ^ (13)
m o 1 o
The above suggests that a plot of 1/U vs In (C./C ) will also give a
straight line with an intercept of value ( - K "/K ') and a slope of 1/K 'L.
° o o o
Hence, the two parameters K ' and K " can be calculated from the plot of
o o
1/U vs In (C./C ).
m 10
It can be seen from Eq. (12) that the capture kinetics are a function of
filter bed, i.e. filter media size and type, bed porosity and tortuosity, flow
geometry, etc.; whilst the second is a variable which modulates the expression
where filtered samples are taken along the length of the filter bed; that
- 10 -
is, Inclusion concentration C, as a function of height, Z is measured.
slope of U /K and passing through the origin. In this approach one need not
to use the first method for melt systems in which only the outlet and inlet
4. PROGRESS TO DATE
diboride system through deep bed filters has been completed. The study has now
been extended so that the performance of a ceramic foam filter may be compared.
The steel filtration study has been continued; a thorough investigation of the
Preliminary filtration runs have also been performed. In tandem with this work
developed. Using experience gained from the first low temperature model study
from melt systems, e.g. Silica from steel melts. Each of these activities is
- 11
4.1 Aluminum Filtration
Using the experience gained during the deep bed filtration stage of the
work the apparatus was modified, so that it would accommodate the ceramic
foam filter, yet still meet the experimental requirements of: sustained heating,
adequate head and metal flow rates in the range of 1 - 17 Kg/(m s) i.e. filter
2
factors of 5 - 90 lbs./in hr. The filter apparatus consisted of a graphite
fixture (Figure 1), in which a two inch thick disc of the ceramic foam filter
material was inserted. A refractory lined steel cylinder was cemented to the
able to enter and exit the assemblies while they were in the furnace, precluding
The size of the orifice at the bottom of the graphite fixture as well as
the metallostatic head within the apparatus regulate the flowrate through each
module. A steel taper placed in the orifice of each module acted as an on/off
The melt to be filtered was prepared exactly as for the deep bed runs
diboride suspension into the filtration apparatus. Once the proper metallostatic
head had been attained the steel taper was rem^oved from the orifice and the
filtered metal was collected in cast iron molds. Flowrates were manipulated
2
to give filter factors in the range of 5 - 90 (Ib/hr-in ). Samples of the
melt system were taken from the transfer crucible prior to filtration, and
outlet samples were taken periodically from the filter apparatus. The flowrate
3
(cm /s) was measured by collecting the filtrate over a specified period of time
and determining its weight. For the flowrates involved in these experiments
18.0 in.) and it is estimated that the metallostatic head varied no more than
- 12
±10% over the course of the experiments.
used in the previously described filtration experiments was used; the only
essential difference being that a smaller filter disc was employed. In contrast
with the other filtration runs, approximately 130 lb. of aluminum was passed
through the foam filter disk in this experiment; while all other procedures
remained the same. All inlet and outlet samples were then chemically and
spectrographically analyzed for TiB^ content. Selected inlet and outlet samples
along with certain infiltrated filter discs were cut, mounted, electropolished
Although the results of this work are still to be fully analyzed, prelimi-
allows one to calculate the kinetic constant K (equation 2), which is a funda-
refractories Al^O^, MgO, Zr02 and Cr^O^ are good candidates for filter material.
13 -
of the filtration process, alumina inclusions were intentionally formed in
the melt in the same manner as was done by Braun et. al (12). Approximately
the first 20 filtration runs were made using electroytic iron as starting
material. The oxygen content of this electrolytic iron was very high -
approximately 1200 to 1800 ppm. Several runs were made to determine the
nature of inclusions formed when cooling the melt in the furnace, or quenching
on a steel plate. In runs (//21 to #27), the melt was contained in an alumina
crucible. Samples from each run were sent out for expert inclusion analysis
by a specialist in the field. The results show the electrolytic iron which
was used in earlier experiments contained oxygen greater than 580 ppm, thus
amount of carbon reduces the initial oxygen level of the melt and further
Using this low oxygen content iron three filtration runs were made. The
oxygen level of the melt was brought up to 400 ppm by appropriate addition of
Fe„0„. Deoxidation was carried out with 0.1% aluminum. Samples of unfiltered
and filtered melt are currently being analyzed. Microscopy of cut samples
from these runs show that there is a significant reduction in inclusion count
analysis is needed.
together with some results from the preliminary scoping filtration experiments.
- 14 -
"^•3 Inclusion trajectory in a converging flow geometry: A study of
the nozzle clogging phenomenon.
due to aluminum deoxidation. Since inclusions are being removed during clogging
then integrated to obtain the trajectory of the inclusion through the tundish
inclusion trajectory within the converging channel as well as the capture mode
of the inclusion in the channel is complementary to the kinetic model which deals
A computer program was written to use the Runga Kutta method of inte-
nozzle. The trajectory which terminates exactly at the end of the nozzle wall
is called the limiting trajectory (see Figure 3). All trajectories to the wall
side of the limiting trajectory contact the nozzle and result in inclusion
trajectory (i.e., towards the nozzle center) pass through unimpeded. Thus,
The effect of the different terms in the force balance equation (force
efficiency was evaluated. The steady state drag force balance is:
- 15 -
Equation (15) is termed the steady state drag force balance because it applies
distance away from any solid boundaries. In equation (15) M is the mass
to a solid boundary was considered. Inclusion motion towards the nozzle wall
is impeded by the added energy required to drain fluid out from between the
inclusion and the nozzle wall. The force balance equation, which takes into
dU _ (. TD —m -1 Cr - (11/2^0
R2p^u|u|
m—
— =
= (m^-m)g - -^^^ ''-—^ ^6)
inclusion from the nozzle wall. Using equation (16) a collection efficiency
of 1.65% was obtained. This is less than the efficiency found previously
in the force balance. The first term to be included is the added mass term,
which accounts for the acceleration of the f;luid adjacent to the inclusion.
dU , , ^./^yp^^Pjl^I'M c, m, dU (17) ■
m ^ = (m^-m)g - kl ^ ^ dl
16 -
where C is the added mass coefficient. When this equation was used in the
computer program it was found that 2.9% of the inclusions passing through the
tundish nozzle are collected. This is an increase over the previous cases
because the added mass term decreases the inclusion's acceleration towards
The second term to be added to the force balance equation is the history
term which takes into account the dependence of the instantaneous drag force
2 I ,
larger than the collection efficiency found in all the preceding cases since
the history term increases the inclusion acceleration towards the nozzle wall.
The above proposed model is compared with the experimental results obtained
4.4 Low Temperature Model System to Simulate the Filtration of Liquid Inclusions
To simulate the removal of liquid inclusions from melt system.s e.g., SiO_
diphenyl in distilled water has been developed. The emulsion has the following
unique properties: (i) the suspension is very stable over several days with
- 17 -
liquid inclusions in a narrow size range of 2 to 4ym; (ii) the density
(iii) the difference in refractive index between water and ditoulyl is high,
Sample ports will be placed at the inlet and after each bed.
the model melt system will be prepared. This will be an emulsion of 0.01%
stabilize droplet size. Scoping experiments with this emulsion have shown
that the droplets range from 2 to 5y; as 3-3 dimethyl-diphenyl has a density,
similar to water, the emulsion formed is stable for several days without the
column and samples will be taken between every 1-2 residence time periods until
the filter appears to be saturated. The samples will then be immediately diluted
with an equal weight of ethanol , in which both of the emulsion components are
the solution into the absorbance range (0-9 p.p.m for ditoulyl) of the spectro-
photometer.
There are three variables in the experiment: velocity,grain size and inter-
of industrial interest) for a combination of grain size (glass bead size) and inter-
facial tension. Flow and logistic considerations would dictate glass beads of 3mm
and 5mm diameter, while their interfacial tension can be varied by baking a film
of silicone oil onto the surface. This changes the glass surface from one not
Ci - %
(19)
Ci
As can be seen from the figure at low melt velocities the efficiency
that the line clearly does not intercept the origin, hence the filter
behavior will fall into the regime where the kinetic parameter, K , is
a function of the melt approach velocity, U^. Figure 6 shows the de-
pendence of KQ on the melt velocity, where it can be seen that the func-
function yields Kg = 0.018 + 0.062 Um, the first term being the value
filter, whereas the second term is the slope and is related to the flow
19 -
off, however, after about 0.40 cm/sec. This would indicate that to
seen that A, the depth of filter required to remove 63% of the incoming
melt velocity.
due to the rapid build up of a filter "cake" on the top surface of the
are presented in Figure 9. It can be seen that the melt approach velocity,
Um) initially remained fairly constant, then abruptly dropping and leveling
off to a value much lower than the initial melt approach velocity. A plot
trate collected over a time period At, and V which is (V-[_ + V2)/2 of each
CK-|_/P)V + (K2/P) for constant pressure cake filtration, i.e., where the
- 20 -
ceramic foam filter disk used in the long term filtration run progressed
from an initial packed bed mode of filtration, where the flow rate re-
8500 cm^ of aluminum had passed through the system. This analysis is
based on the fact that as the cake forms and builds up, the pressure (AP)
required to force metal through the cake increases; however, since this
creasing flow of metal through the system, which was in fact observed.
the long term filtration run clearly showed that a cake had formed on
the top surface. Although inlet and outlet samples were taken during
volume of the filter because: (i) the filter enters a cake filtration
mode which is not encompassed by the proposed kinetic model; (ii) the
samples are analyzed only for titani'om diboride content, hence there is
in the melt.
this variable in all the experiments has been minimized by using a con-
21 -
sistent titanium diboride master alloy, systematic stirring, etc.
alloy was used as the starting material, rather than pure iron. The
following contributions:
22 -
5.3 Inclusion Trajectory in a Converging Flow Geometry
(12)
The proposed model may be compared to Singh's experimental
results. Singh reported that flow through the nozzle shown in Figure
of 21 cm^ which is more than enough to reduce the flow through Singh's
nozzle exit. Thus, the total volume of the agglomeration is much more
- 23 -
6. FUTURE WORK
• The performance of the deep bed and ceramic foam filters will
be compared and contrasted with reference to the kinetic model.
6 .3 Superalloy ^Filtration
- 24
REFERENCES
8. J.J. Farber, F.M. Rupp, Light Metals, Vol. 2 (1979), pp. 733-740.
9. K.J. Brondyke, P.D. Hess, Trans. TMS-AIME (1964), Vol. 230, pp.
1553-1556.
10. M.V. Brant, B.C. Bone and E.F. Emley, Journal of Metals, March (1971)
Vol. 23, pp. 48-53.
12. T.B. Braun, J.F. Elliot, and M.C. Flemings: Met. Trans. Vol. lOB,
June 1979, pp. 171-184.
- 25 -
TABLE 1
Ci-Co
Sample UmCcm/s) 1/Um Cl(ppm) Co(ppm) ''^- Cl
ST201 0.22 4.53 979.2 382.5 0.61
n
S-i
3
OJ
en
•H
3
•H
OJ
01
O
<4H
3
3
•H
O
3
X
en
01
U
3
O
3
o
o
>
O
•H
o
C
en
X3
en
>
3
4J
QJ
01
(-1
■H
3
4-1
CJ
3
0)
-i3
1
>?
c3
•H
SH
4-1
CU
Melted at 1600°C, melt is held Duplex inclusion SEM study
m
%
^ -H
C X
rH
flH -H
w -H
O 3
u r-i
3
O B
3
0)
OJ 3
rH 3
for 30 min. at 1600°C, solid shows the presence of Fe,
>, 3 6
O
U ^
•H 3 <U
U 0 U
O -H OJ
O
a
_^
3
4-1
CO
ca
M
•
aluminum (0.09 wt%) is dropped Al and 0 in inclusions.
o a A
SO
in the melt. The induction Size of inclusion: 2-40ym.
power was shut off after 5 min. Near crucible wall clusters
of aluminum addition. of inclusions were present.
Spherical inclusions of
CM
(N
Electrolytic iron melted No Al or Fe203 was added. The
at 1600°C in alumina sample was held as 1600°C for size 2-10 pm were present.
crucible in argon 11/2 hrs, and then furnace SEM study shows the presence
atmosphere. cooled. of Fe and 0 in inclusions.
eg
Electrolytic iron + Melted in vacuum ('^ 10 microns) The sample was very clean.
0.0 25 % C in alumina and then furnace cooled after Negligible amount of small
crucible in argon holding at 1600°C for 30 min. inclusions 1 to 5 ym were
atmosphere. present.
TABLE 3
Wt % AI2C 3 Wt % Oxygen
Run No. Inlet to filter Outlet Inlet to filter Outlet
28
29
analysis unavailable at the time of ^inriting this report .
30
31
Al > lK)CLU5\Ok)
STEEL TUBE
RESISTANCE
FURNACE
GRAPH\TE FlXTliRE
7 ■^'
5ELEE RLTER
^-^-^
TT
FILTRATE
/
M0UDC5)
10 cm
5.5
5.0
4.5
E
-^ 4.0 —
3.5
3.0
2.5
Figure 3,
Limiting trajectory computed from the model (equation 9) shown at'
the exit side of the converging nozzle. Z indicates the height
from the nozzle exit, and R is the radial distance from the lozzie
center.
1.0
o.a
a6
I a4 -•
cxx ••
U^C-id'^-^)
c - c
Figure 4. TiB2 inclusion removal efficeincy, n = _J; o , as a function
of tke melt approach velocity, Umo i
3 -■
^ ■■
U(%)
U„ (Vc-)
velocity, U .
ao7--
0.0* ■•
o.os ■•
K.
aca ■•
0.01 ■•
aoi -■
0,^00 ■ \
A \
A
0.2SO ■ • \
X CcV) \
0.Z0O -
\ A
^ A
O.ISO - ■ d
A
-^.^ A
0.100 - ■
A
0.05O — 11 t > 1
10 20 30 40 50 60 70 80
U.
'M
Mo"'-?)
K
figure 7. A, i.e. T^,asa function of the melt approach, velocity, Xl^.
U
m
Figure 8. Depth of ceramic foam required to remove 63% of TiB2 inclusions,
A, as a function of the melt approach velocity, U^.
•
(•i6'^-^)
as ..
At {5/C^)
AV
o.i ..
O.I •.
140C
VOLUME (cW), V
No. of
Copies To
Commander, U.S. Army Troop Support and Aviation Materiel Readiness Command,
4300 Goodfellow Boulevard, St. Louis, Missouri 63120
1 ATTN: DRSTS, Mr. J. Vfurphy
' Commander, U.S. Army Mobility Equipment Research and Development Command,
Fort Belvoir, Virginia 22060
2 ATTN: Technical Documents Center, Building 315
Cabot Corporation, Machinery Division, P.O. Box 1101, Pampa, Texas 79065
1 -ATTN: W." L. Hallerberg, Director of .Metallurgy
6 u «
Q 1^ « m
S'? a5 .2 -
u « « 4 U
' m Q « « )
I a ka U k
« 3 f^ • •
O
— • a
« 3
—
■ £
M
*>
« Q
»*
«
"
a ■ a « 3 a. *
44 3 ~> h. ■ u 3
■O ** « -^ O >j 4 a «
c =. « ■« u
i B 4J 3 B W K — a TS 3
^ C7- 7 « 3. ^
^ .. _ u 3 3 -5
V «
w -- a a
• i (^ « -^
Bk, _J < VI VI Cl. w fj a - "a fl ' « a -^d
■ t,. >s O -^ -3 « -H ■ ■
3 ^ O • a w X o
■ W U <M
* S w ■ -• 3 .C O
■a -I « -«
«l « u • •«
• ^ W V -^
•H w w u « PH
e 4j e « a • v4 w hi u 4) •H
O ^ Q 3 « a «
'>' -^ •> U « B u e « « •
« o 3 • -^ -g o a o s »^
ta ■- -1 3 -O C
-- -^4 « U •
W kJ u -4 4
w e a M a
4 O 3 4 -t -^
w u u ^ «
>w w
>M -* a o -q -^
V « 3
s. rt 4 C fl -g
IM IJ « « O
a ^ u •
■ » -- -3 -4 • e
u « a «> !■ a. O
-H a >v -^ u 3 ■»* a o -H -a -^ • fl
U 2.-4 S « « W Ej a a D ki a. o
S ! S -•
u
■» >. "O u
a. -4 c 9»
3 -^
a M
Jt 3 V ■ -^
« a g « « 4
c o c tJ -a c t*
4. - -• ■*' 5 — u a X — -^
i> -o r3 3^ ■3 « C W 3 -4 3 u a -^
3 = = ■-)» C O •
^ 3 j -* « O 3 -^ o\ o ^ c
C O I S
M
3 a
3
11 i
-H *T 3 -
o > ■-
«N o u
Z • o 3 a r 11
— ■IX.
25 ?
o -^ 3^ < Z • o 3 a S 1 V ^ a a e >
-- H TJ -. - -j ^ 1 CQ u W fl u -4 o
(H fl —. -H fl j= J: •.* o -^
S5 5 i- <
r^ k. -•
t ^ <r
03
a>
^ 3 a -3 ** o -g
•J 3 s
0^ " 5 "
j fl u Tj -• i^ C O 4 ^ 3 a -3 ** o "3
a ^ « e - V « o *J u 3 a
•-• ;jB 31 n U < >« « > -" o -j -3 -. u c o a
ec > a: «-> i- « O »-< • « q a c w a o M
^3 I c — ^ - O "^ X ^
« -j
= a
-.
3 -O -^
-t g u
«
a > -4
O <^ a • o
a
J =* ■= u c ac — ■£) -5
^ < w u a 4
C fl
o M
t c «
a -4 « -*
fl « ^ k. X Tj c sa 3-^-3
0-3 -H M u •< H u i a
1 o 3: c -' , c a IS •*
Q w a >H a -*
0.3 -H ta
■a ji a C -3 •
c n >. -w ^ a 4*
• --:i!-*.--i--v } ka 4 « "H I
3 -4 3 s 3 a o
o M -4 3 a* s a
■
a u I u ^ .^ -^ w Id
^'^i°:3i5'K-—"^^ - V « i« « se 4< 4 a 'H u a
aS-i^* T"a
_i.!iita-H-3-« .M-*
«
u
3.
3
J= O 1
H B
.!< u u u ^
w =.iy«(SM—c-i. 3 M >< fl .^ a. X o a
M O -
ifc-<_»<-.va. w
^ ^3 .'3i-^fl^--^C
-, -3 3 a
3XOtjaeux-»Q« 1 ^ u _ —
.1 I
5 1
r z a u a 3, X -.
11
»!
w a
8 ^a
m m u
ss 1 s.
■n\\ XV a. m
3^
a Wa
x a
a §
o ^
^ u a »
W 3 4J O
S «
a
■
M
ilo
K ti e ^ M a c *j o ^
a _ u a a •< u
a a w 1
a a "fl a o U )ri u
a e ki >> "4 a 3 a. • a a O a a o a 3 3. > a
a -3 3 *J 3 '-' i -* u ^
- e s — a w « a N e u o -* X 1-1 ti **
k. a u 3 s s — m .4 a O « 6
a ^ c fl -H •3 « -" -H 0 u a b> a
c a. a -t o ■3 a ^ -^ O
-j 3 e * « -* a a *. c a. a -4 «J
-H 3* 3 C C ka a a > a a *<
-.4 -• -^ a-l 3 3 -3 w -. ■ e _• cr 3 a a. ka u a >
BH _J < V) V) &• *•
a u >v
•
o
.5
'-'
-* -a _ t. 3 3
Cb ..J < VI WI b
•Q
a
44 -M
fl
a a
o
a w >, o -S
3^0 • ■ a ii %i w
u sa w w M 3 X o a
U OS IJ U hi
" X « C -H fl -M U -H *J
■o —.a ** -* X a a -^
a a w > »• •o —> « -*
a i u a -0 a a u ■ -M
^ u 4> u a ^ a X M o •«
a a ■H k* lai U a iH
e w s a a a a a
o a o 3 w s u e a a •
-. -* • y a o a 3 3 w
aj C a iJ a •4 -^ a u a
w e a u a
:5^u -•3?"aa
aj, «
a o 3 a -H -g
o -H a c a -g aA •■> u -4 a
O — a a a -j
<H w a a o
<D
«N i«
•sS'^s-s^ a ^ M • 'sH'" S-31
tr\ o a a -* T3 -J a B
fi
O F-H u a a a 1- a. o
o <^*
: S.i-S S 3 2 ^ ■-
w
fl >.-g o
CL-^ c a
o o
S ! S2-S a 3 u <« a
c
B « a a
Q c X n ::- 3
* ■ ^ "i
*9t -1 u a >* -■* -^ u a
(O 1 WD
-^ = a >
a a -4
,•■: =--;3
c o w V e o a
Sa 8a K ) V to
o^ ~o -0 -^ a a .. 3
M -T 3 •«
U 'J a « -1 a o > -^ U B 2 o -.
a « -i a o > -3
*j j -« >
s-s a.:; ?. S.
O
— -1
Or
3 d X
-3a
tj i; <
"
fl .j « .* 1- ■^ 0
•4 fl X £ «J 3 r-
u X • O 3 O
-a -1
X
y
"
G.
s's i »^ ?*
a aj V -a i-> •< A
fl N O U
•-• - -j < 1 c -a J -H ^iM a r^ -« ■" fl X X *- O a^
C^ fl -.
X 3 a -3 ** 0 -g
a -■» H -g S O-SSt
fl —*
c -3 <-■ -' a*-^
fl Q. fl >^ •- C
X 3 a -3 •« o *s
— —I c m -4
e go s =
u -^ -4 4» B O • -sg'
a ^ e
a
■^
I
-.
- o- cT>
c » -t
O 3 fl
a U < >> u a •H a e « Q = e £ a a o w
<T4 *> U. 01 a LJ < >< a > -^ fl
X -D C BO <M} > a: >^ i- a 3 «* a a b
:2 ;. .j 5
^ < H « ■ a
.H ^ fl .
ct i-j
a a <
a
I. C ^
>s
^
1
• o a
m -< 3
X -a c M 3 -o -*
** 5 -* -( a .-
■-*
c a L = ■*
X a a
*j ■< w -I a a
c S - ^ « U 3 3- X
c 5 — -o fl
z _2 a >s r o» -? "1 ~ 3.3 *4 W
U X 1^
>. = S .c - -)
11
2*5 ::i:3
i: 5 23 = j •3 .a a C "^ •
X u
fl
5=
O
3
X
0.3
■fl X a c -a
-^ w
■a : c a ?-.— -^ • **
c a >, -H -* a -
•
= -"
o '3
«i ^4 "3 cr
H. a
ii y
a
a a* u « " 2
> -4 o 3 " a- £U WKfli-a-G
u L. fl a -H w a a — - >-.03>-'a-
^ u M a ^ fl a -4 k> a
a c - u a ^
— 3 t
u X a a *- *- u Jc— floc— 5j— a«
o a fl u a-js o • , a a lu <u u
k. O - Q.X o a
3 3 :L^ ^ u *■ I o o
fl ex Oa U irftoaai:*— T3^ -k.« 3 R ■< H
fl a ( uu a.waflaci.ekau 3 a '^ "a
jS 1 - •
0
■ 3^ i a
i*
■« a ti 0 w w i 3 • U
M
V
u n ■ a4
« C -4
«
U
3
u •a *-t
3
- S'S > •
C3
u
Zm
a c u 0
c °ss ;. « b M V C ^ t*
3
a
C *J
o »
« o
>. -<
o <-»
•
X
3 & •
M V>
«
W
>M « ^
a
l-r
■
c
a
O
a
w « ■ <
V «
a
w
1
-S a a - 'n ■1 0 "m a L. Ill
V « 3 - J
u « 0 «
L. a u
fl
3 <a
0
:x c « >>
0
a 3 & . t
n T) n n -* ■a « -, ^
c 0. « -*
a a -^
0
u
a ■*.
>-< 0 «
-4 s a
0
-
•J
•n ^ a 0 a
M
e
** •
*j 3 a « « — « 3 -H Wa a U 3
-^ tr :: « Q. M c — 1^ a -0 c 19 ■73 V -1 ~* 0
_j -, _ iJ 3 3 -0 4j — a ■ e 3 a c a. fl -M
n, ^ < (A (/I cw •^ 3 & V w as <M
— cr 3 w a. S m
« « 1.
3 £ 0
U M u U
«M iJ
a
u
=)a fc ZJ < 1^ «
3
0.
13 - -t ■
V a
c
• -4 U -4 *i
-H J: «
■a —^
« a
a
a
u
w
-^
••«
lisZ a
■ J: M 41 -Q
^ M u
B 4J
u « -H
C a a «
a « 'a X.v:
w u
■4
0 a 0 3 >^
•4
M e
-H
«
•
w
u
a
a
c - c a : t
3 U
;3^ 351 u C « u
u a
M
-. -2
O
o■«■
^Hil-a'
*- *j V a o
a g 3 a
w -* -^ 3
U « (J ^ a
-1 a c a
a>
s <M ■- -« k. .^ > ■? „
i
0 0-0-4
0 ^ *i M a <M u a a a
w «rt a ^ K •
trt o « a -H -3 -H a 0 <*- "H 0 0 ■fl -H
O -* u a> « a) k. Q. 0 a i« w V a
O c** a
u <o
u ^ -t c a a w S»0
S!u S2^
-« a X
= !
-1 ■-
- (J«
^ s « ■ -d U 4
4. -. _ -> -a a c 3 c: 1-1 ^ e M
»> -o -rj ^^ -3 M C k4 3
-H -^ o —t c 3 u V c 0 a
a w -^ « 0 > -H
(." CJ
u ja -w > ■3
C
••5 =-i;3
0 u u c 0 a
O 5 a X t- g 3 3 « « c j a « -4 m 0 > -rt
w -1 '^ a. .J X -t >
1^ 3 .S?,
— fl X j= *• 0 --
2 -SSf C T3 ^ ^
■4 o. a >, -^
M-M
c
O 3 2 X :
*j V fl k. -< 0
ffi (C -^
X 3 a T> >w 0 "5 flX X u 0 -^
■ 1 u ^ a< 2 ."5-
« c -0 u — «*«
-« ^ i 03 — si -a -. 3 c 0 a
a c » • 0 *J _ «a o sIM
4 _^* 2 3.
3
« >v -H e
a -0 •* 0 ■«
^ (a=
C SO O =
« > -" „ a a 1 3 C
> [£ </^ - u 0 f^ a V q a c -a -t *- e 0 a
u <
c£
c <o s
a t C — (-> • O f^ X -0
*" - -< H u 3 a
c oa 3 -o -3
-1
a
ba
oe > OS
0
v^
c u « 0 <J
a > — s
0
-aS^S^3
c
a CJ 11 f^
c 2 — ^c
=
■^ c
0
a
w
1 c
a -4
«
» ■-< X
a
a> w3
i: —
>.
Ji
H
U
1
a X
2<?:t 3;;N
■0
c
J3 a c
fl >N -w
-a
-^ a w
a
X z >. !- X
yO c
a
a L c
s<- M
9C
c
41 « I- » -- a 0
U
3 -H 3 a
w ^ 3
^ ,^
3
tr
-H
0 y
c a
W bi
Tj
c
a
e
0 X c
aa
a. <
c
0
-3 X a
• r
c -a
-H ta
a
1
> — O 3 -J tl. H
Qo w « a -^ i-> a a oC 3 -. 3 a 3 « u
o C D. 4; — 01 a: j<
<J
4i u
J= « w
ax
»*- *- U
U
>X c
s
0 u ^
^
3 a- s a
^
V ^ "a * T '0 a
u s n S.U. k.
■s
CQ
u a.x
U 3 H
3 «
0 «
>N
of )i
a
a
-1
C
■y a:
c
a 0 0-
-« K
01
■J
■e
k.
V
a
w
« ^ k. a
ax
Q =j ^ o-x 0 a
'Z, «-)-.' < jj
:^ 1 3 V
< «
4( ■a
0.
T3 —
^ T
u
3
3 H
V >.
a£ k<
V
(»< a<
a
>,3Xa<jQa-x-
2 •i
a dz 0
•■■,
a e
1
w
^
«•
3 Q u a a- X ■« a a ■**
<fi %**
•
a a ^
1: a Q ^
a:
s :;
■ n
X ""
JTJ
w hi
a
i tS^ :9 XV
il .2
Oi
V e 4j o
X a e ^
a o
« 3 O.
0 «
— SB - a o
-J 3 —
a -O C « - laa
a i-i 3
13 «
^ -f o
c a.
a -^ u
a a■-■
-3 u -H B aa •
• -J < VI V? b « a -« w
-M a s
a w >, 3 u a 0
a fc. X o -*
a a V4 Hrf w
3x0 a
e M e • '
w e a w ■
o a o
*4 -H V
a o 3 a -H -5 u e a w
W — r-t 3 -o c a o 3 a •
M ii y -^ a
-1 a c a "S
<M u a a o •sa „o
*- *« o o -a <-i
o •« w V a >t- — o o
a a -< ^ -1 a B
q *« »j ■2^
a a -S -n
u w, » *i
u &-H e a a <■* -4 a X -a
u s a « a a -i C-' c
c o 5 " -o o •« V e < u
■rt u a X -^ ■-' c o e "
-i u a <
Jit C J
a -4
a C
a 4< -• V
O 3 E. i: 1- 3 V 3 a a a >
a u M a fc" -^ o
■H -^ X X *■» o -*
C 13 hJ -^ M'M
|-s it
a u « a
..1
a
BO
a a
— 1 X X
a a. a >. -^ fl — i*>M
X 3 a 13 ««- o T»
u 3 o
■
u*— a
_JC -4-<Ca5-4
(iicox
!U -g — w c o a
H c V a O u
•x ^ C 3D ^
!- °1
• c o a
a -IS
C es o X
V Q <s a a q
.« U •< X
X -o c « 3 -o -fl ■* a a q
I.J W -.^ rt a **
«X "-J* CM — ^(« C a
w < H
[
u a a
C 3
;s ;;i -o a
) -I 3
- < H
i a -o -^
*» 1 a
XuZ3a>,c:rx-<-i o u a -rH a ■-« C M
o
^ X a c
• 3
^
"Hi*
a
^ 5 £ " :S J! O U fa ^ a «4
O • 3
"3 'EH ."i'H^'^T'*-' c a >,-*
3 -t o a
-H a u
3 U U
■o J3 a c
3 ti .x 3 9 a a
•§ SM c a X -<
3 —< o fl
«
„
-.- >-.03"S.
.U.C**-.XO.-.u*
^ ^ -4 -4 W ki
-- sj ':33".5 :; % o »» -^ 3
a c-iaoccw—iwoc JL V \. _ __ u. c <*a — X a- -
U X V U <H -H t>
:six i-'"Ts»
aJU i^v««i«a.b.i..
a aj
oa u
a
<xx o a
3 P
a >%
; o a
a< M
* 8 2
ec —
:=> ^ a
IJ -1 ec
iJ
-1
t
—)
O B
■3
3'
3 a
.■3 1
^.
>% *T
_« 'n^
Jk« ^\
L-l 1k*J\ ^
tJ &.
•*• " .^
« ^ >4
"^ tJ '^ • O *- X c «^ -^ a
k u a &• X -c a •-•
1 3
1 iJ