LANTS

I
I

Filtration-Extraction of Cottonseed
E . L. D'AQUIN, H. L. E. VIX, J. J . SPADARO, A. V. GRACI, JR., P. H. EAVES, C. G. REUTHER, JR., L. J. MOLAISON,
E. J. McCOURTNEY, A. J. CROVETTO, AND E. A. GASTROCK
Southern Regional Research Laboratory, New Orleans, La.

N. B. KNOEPFLER, National Cottonseed Producfs Association, Dallas, rex.

T HE development of a satisfactory, simplified process for

direct solvent extraction of cottonseed, yielding meal and oil
of high quality, is definitely desired by the cottonseed industry,
filtration. Also indicated was a high tonnage capacity per
square foot of filter area.
These oonsiderations pointed t o the selection of the standard-
particularly by the small- and medium-sized mills, before there is type, continuous, horizontal, rotary vacuum filter in preference t o
further expansion of conventional processes. the disk and drum types (10, 11, IS). Also, the horizontal type
Direct solvent extraction in immersion-type extractors pre- permits efficient countercurrent, displacement washing, efficient
sents a serious fines problem ( 1 , I%, 16). Moreover, the main- drainage of the original slurry and of the cake after each stage of
tenance of consistently low residual lipides in the extracted multiple washing, allowing minimum filtrate holdup, and pro-
flakes has proved difficult (14). Percolation extractors, such as vides for continuously clearing the screen. In addition, the
the basket, cell, and band types, present the problem of obtaining final spent marc (solvent-damp cake) was of the nonsticky and
relatively rapid and efficient percolation of solvent and oil-solvent free-flowing type that could be continuously discharged by
mixtures through thick material beds. means of a scroll.
The pilot plant development of a simplified solvent process, The use of a horizontal vacuum filter t o conduct counter-
called filtration-extraction (8),is described here; it apparently current washing is not a new or radical application. However,
overcomes most of the usual problems with direct solvent ex- the combined operations of extracting a solid material for a de-
traction and is applicable under the conditions met in the smaller sirable component and subsequent countercurrent washing is not
aa well as in larger mills. The principal unit of equipment is a a usual commercial practice and, in fact, is an innovation in oil-
continuous, horizontal vacuum filter. The oil extraction step seeds extraction technology.
consists of a flake-miscella mixing-soaking operation followed It was necessary, before accurate performance data on the
by high-capacity vacuum filtration and countercurrent washing. pilot plant filter could be obtained, t o determine the respective
Another departure of the process from conventional solvent ex- screen a r e a required for best placement of slurry feed and washes
traction is that i t embodies a mild precooking of the flakes. and for drainage of the initial slurry and of the cake after each of
Development of the process on a pilot plant scale is based on a the washes. Vacuum and blowback requirements were then
batch extraction made in the laboratory (18). The data re- established, and observations were made relative t o cover
ported here supplement pilot plant data reported previously blinding, blowback performance, and cake discharge over ex-
( 7 4 , which included some preliminary runs with cottonseed. tended periods.
The filter (Figure 1) used in the work reported here has been
Process Combines Soaking to Extract Oil described previously (8). The only alteration from standard is
and Horizontal Vacuum Filter to Remove Miscella enclosure in an adequately vapor-tight hood. The filter is of all-
iron construction, and consists essentially of an annular, rotating
The fast settling characteristics of the slurry of miscella (oil- pan, containing wedge-shaped screen sections, each of which has
solvent mixture) and cooked flakes demonstrated in the pre- outlet ports t o a common centrally located multiport valve below
pilot plant studies indicated that larger scale filtration and wash- the pan. The valve is so constructed as to lend itself t o a variety
ing should be conducted on a horizontal plane and that vacuum of filtrate separations, which makes i t particularly adaptable t o
should be used in preference to pressure. Also, the earlier studies countercurrent washings. A typical combination of the valve
indicated t h a t the filter medium should be solid, monofilamental filtrate outlets t o attain the countercurrent washing procedure
material, such as metal or saran, rather than multifilament yarns, for the four separations used in this process is shown in Figure 2.
such as cotton ducks. The filter pan is turned by a variable-speed drive between 1
I n the previous work rapid filtration (high mass velocities) of and 3 minutes per revolution. The scroll which discharges the
the initial slurry and of the three countercurrent washes was ob- cake from the pan is shown in Figure 3. It is of triple-flight,
tained with Et horizontal screen and a filter cake up t o 4 inches standard pitch construction, and revolves a t 70 r.p.m.
thick and under vacuum of less than 5 inches of mercury. These The filter medium used for the initial pilot plant runs was a
results indicated t h a t the cake was satisfactory with regard t o 20 X 250 mesh, stainless steel Dutch twill wire screen. This was
particle-size distribution, porosity, noncompressibility, and other later replaced by a plain Dutchscreenof 24 X IlOmesh, whichcosts
considerations given in the literature (2, 4, 10) and that its less, is more sturdy, and permits satisfactory fines retention and
specific resistance was considerably below the minimum of faster filtration rates than with the finer-mesh screen.
5 x 1011 feet per pound (a cake 1/8 inch thick can be formed in 5 B y means of blowback nitrogen gas, introduced continuously
minutes or less a t a vacuum of 20 inches of mercury), which at a point immediately underneath the filter section onto which
is recommended by Grace (10) for satisfactory industrial vacuum the slurry is deposited, the oncoming slurry is mixed with the
241
248 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 45, No. 1

The slurry is transferred t o the filter by means of an inclined,

3-inch tubular screw conveyor (Figure 3, top left), powered by a
hydraulic-type variable-speed drive. Four 10-gallon vertical
receivers are mounted below the filter t o collect the separate
filtrates from the respective outlets of the multiport filter valve.
They are connected a t the top t o a header through which vacuum
is applied by means of a wet-seal pump, which uses hexane as the
sealing liquid. -4separate centrifugal pump handles the filtrate
from each receiver. This asscmbly is shown in Figure 5 .
The system for application of continuous blowback gas com-
prises a nitrogen gas cylinder, a variable pressure reducing valve, A
rotameter, and a pressure gage.
&4ll equipment units are rlectrically grounded. Electrical
equipment is Class I, Group D.
The pressure filters used for polishing the miscella, the con-
tinuous evaporators, and steam-stripping column for oil recovery,
and the continuous screw-type desolventizers for meal recovery
have been previously described (6).
Conversion of the continuous solvent-extraction pilot plant a t
the Southern Regional Research Laboratory t o filtration-extrac-
tion was relatively simple. It required merely replacing the
scrcw-type immersion extractor by the filter, with mixing con-
veyor and accessories.

Flakes A r e Cooked M i l d l y
and Cooled for Extraction
Cottonseed of three different histories was used: Current prime
seed (A) with a free fatty acid content of 0.53%, from Green-
wood, Miss., which was received and processed in November 1951;
current prime seed (B) with a free fatty acid of 0.80%, also from
Greenwood, Miss. , received and processed in January 1952;
and current subquality seed (C) with a free fatty acid of 5.2%,
which was received from Alagnolia, hliss., and processed in
November 1951.
Figure 1. Continuous Rotary Horizontal Pilot Plant Filter,
%Foot Diameter f SCROLL

layer of meal (1/16 t o 1/8 inch thick) passing underneath the dis-
charge scroll. This operation helps to keep the filter medium
clean. Vacuum is applied t o the system through a common
header connecting the four filtrate receivers, described later.

Mixing Conveyor .and Filter Convert Plant

from Solvent-Extraction to Filtration-Extraction
The hullers, meats separator shaker screens, purifier, hull
beater, flaking rolls, and five-high cooker which were used for the
material preparation are small-size, conventional-type, such as
currently used in cottonseed mills.
The five-high, steam-jacketed cooker is equipped with vents,
atomizing sprays, steam ejectors, and indicating and recording
instruments t o control as closely as possible the moisture and
temperature of the flakes in each cooker ring. Open drying
trays constructed of '/a-inch mesh wire are used t o cool the cooked
flakes. Figure 9. Arrangement of M u l t i p o r t V a l v e Outlets in Horizontal
A simple, paddle-type mixing conveyor (Figure 4) was con- Rotary Filter
structed in which t o mix thoroughly the slurry of cooked flakes
and miscella. It consists of a vapor-tight, standard U-trough, 1 The material was prepared for extraction under the conditions
foot in diameter by 12 feet long, having a shaft with adjustable indicated in Table I. The cottonseed was first cleaned and de-
flat paddles set at 7 degrees t o the centerline. The shaft is linted. The meats preparation equipment was operated so as t o
turned by a l/*-hp. hydraulic-type variable-speed drive, capable of produce a mixture of whole and fine meats having a moisture con-
any speed between 0 and 25 r.p.m. The speed used depends on tent ranging from 6.8 t o 8.7% and containing some hulls. I n none
the filter feed rate, the angle of paddles, and the slurry character- of these runs was moisture added t o meats t o improve flake
istics. formation ( 1 7 )or t o flakes prior to cooking.
A screw-type feeder ( 6 ) ,built in the laboratory shops, 4 inches T o prepare flakes for the first run the meats were rolled through
in diameter, with variable-speed drive and plug-type vapor seal corrugated rolls (one pair high) spaced a t 0.016 inch and then
mechanism feeds the cooked material t o the mixing conveyor. through smooth rolls (one pair high) set t o produce flakes 0.004 t o
January 1953 INDUSTRIAL AND ENGINEERING CHEMISTRY 249

The cooked material is fed t o the continuous mixing conveyor

Table I. Material Preparation Data where it is slurried with the No. 2 filtrate under gentle agitation
Run No. 1 2 3 4 5 for 15 t o 20 minutes. During this soaking period the greater
Cottonseed A B B B C part of the oil goes into solution as concentrated miscella. Coats
Free fatty acid, yo 0.53 0.80 0.80 0.80 5.20 and Karnofsky (3) have shown that the rate of solution of the oil
Lb. processed 1600 1600 1000 1400 1300
Meatsfraction. lb. 1010 1052 619 891 738 from raw cottonseed flakes soaked in hexane miscella is relatively
Meats fraction, mois-
6.8 8.2 8.6 8 7 7.2 independent of the concentration of the extracting solution and is
Flaktt?io?fs type l-Higha 5-High 5-High 5-High 5-High mainly a matter of soaking time. The same is true of cooked,
Flake feed' rate to
cooker, lb./hr. 260 260 260 260 260 re-rolled flakes, but the action is significantly more rapid and
Cooking conditions complete.
Total time, min. 60 60 60 60 60
Temperature, O F. The slurry is continuously deposited on the revolving pan of
Ring: 1st 208 202 203 200 210 the filter, where, in from 1 t o 2 minutes, the bulk of the concen-
2nd 215 211 211 214 215
3rd 216 214 214 218 218 trated, or full, miscella is drained off by vacuum and the re-
4th 213 217 217 219 218
5th 223 222 222 223 223 mainder, with some additional oil, is extracted by the use of
Moisture, 7' three successive countercurrent washes and drains, the first,
Ring: 1st 25.8 15.5 17.5 18.2 15.4
5th 12.7 8.4 11.8 8.9 6.4 two with progressively weaker miscella and the third with
Cooling and reforming hexane. The initial or most concentrated filtrate, which is drawn
conditions
Time cooled. min. 15 10 10 15 12 through the first three outlets of the filter valve to a receiver be-
Temp. after cooling, low, is the final miscella and is continuously pumped to the oil
F. 135 135 135 140 140
Rolls for reforming, recovery system.
type 5-Highb I-High0 1-HighC None None
The deposited cake, after the initial drain, moves forward onto
a 1 pair high cracking, 1 pair high flaking.
b Two bottom rolls only. a blanked-off outlet directly below the first spray, or wash,
e 1 pair high flaking. which is allowed t o pool before it is drawn through the next two
valve outlets t o the second receiver. This filtrate is the second
most concentrated miscella and is pumped t o the mixing con-
0.007 iach thick. For the remaining four runs the meats were veyor for slurrying. Similarly, the second wash is applied,
flaked through five-high rolls (top two corrugated) set t o produce pooled, and drained, and the filtrate of this wash is used as the
flakes in the same thickness range. first wash. The third and last wash is hexane, the filtrate of
For all five runs, flakes were fed to the cooker at a rate of 260 which is used as the second wash. After the final wash is drained
ounds per hour and cooked for 12 minutes in each of the five off, the fast-moving scroll discharges the marc into a chute which
fettles-a total cooking time of I hour. Cooking conditions empties into the desolventizers below.
were comparable except for the moisture added in the first ring
for run 1, 25.8'%, as against 15 t o 18% for the other four runs.
The purpose in adjusting the moisture content t o a relatively high Filter Operates with 1-Inch Cake, Pan Speed of
level is t o accomplish moist cooking as the temperature is raised 1.6 Min./Rev., and Vacuum of 1.5 to 6 Inches H g
t o about 210' to 215' F. in the second ring which facilitates
binding of the bulk of the free gossypol with the protein compo- Operation data for five pilot-plant runs are included in Table
nents of the meal (17'). I n the first ring, the temperatures for the 11. The operating time per run, which ranged between 2 and 3
runs ranged from 200" to 210" F. I n successive rings, tempera- hours, represents the actual time the filter was in operation to
tures were gradually increased t o about 223" F. in the last ring.
Moisture contents were reduced during cooking t o a range of 6.4 process the specified pounds of material.
-- --..
tn 12 7 %
The &,-cooked flakes were cooled from about 223" to 140' F.
by spreading on the open trays for 10 to 20 minutes. T h e mois-
ture content decreased as much as 2oJ, during this operation. Table II. Filtration-Extraction Data
The batch scale studies had shown that the best results with Run No. 1 2 3 4 5
filtration-extraction were obtained with cooked material. A Cooked material processed, lb. 910 709 509 525 575
secondary reason for cooking, particularly by the techniques de- Feed rate, lb./hr. 300 300 300 300 300
scribed, is t o produce meal of improved nutritional quality. It Moisture, % 10.1 10.4 9.9 8.8 7 6
Lipides, % 27.2 28.1 29.0 29.9 31 7
had also been indicated t h a t the cooked flakes should be re-rolled Solvent-flake ratio 1.1:1 1.1:l 1.1:l 1.1:l 1 1:l
after cooling. Re-rolling flattens out small agglomerates (Figure Slurry temperature, F. 83 82 91 87
6) which may form in the cooker, caused by hulls curling around Solids in slurry, % 3i:3 31.1 28.0 25.7 28 0
Filter pan speed, min./rev. 1.9 1.4 1.4 1.4 1 4
meats particles and hindering efficient extraction. T h e cooled Cake thickness, inches 0.8 1.0 1 .o 1.0 1 0
flakes were re-rolled for runs 1 t o 3, using rolls (one pair high) Wash rates, gal./min. 0.9 0.9 0.9 1.0 0.9
except in run 1, where only the bottom smooth rolls of the five- Vacuum required, inches mer-
cury 6.0 6.0 3.5 3.0 1 5
high set were used. Lipides in filtrate, %
1st 26.0 22.1 24.5 25.3 25 8
The combination of these operations resulted in t h e production 2nd 9.7 9.1 11.8 11.6 10 7
3rd 3.7 0.7 2.1 1.3 1 5
of crisp, granularlike material from which t h e oil could readily be 4th 0.5 0.4 0.6 0.5 0 3
Pines in first filtrate. % 0.13 0.25 0.17 0.21 0 30
extracted. Solvent in marc, % ' - 27.8 22.1 21.3 28.4 24 8
Bulk densities were as follows: Flakes t o the cooker, 22 pounds Residual lipides in desolvent-
ized meal yo 0.90 0.66 0.83 1.02 1 49
per cubic foot unpacked and 27 pounds packed; cooled cooked
flakes t o the slurrying conveyor, 36 pounds unpacked and 47
pounds packed; final desolventized extracted meal, 42 pounds un-
packed and 46 pounds packed. The prepared material with a n oil content by weight of 27 t o
Samples of the following were taken for each run: cottonseed, 32'% was fed t o the mixing conveyor trough at a rate of 300
meats before and after flaking, flakes in top ring, flakes after pounds per hour. The second filtrate miscella, containing about
cooking, after cooling, and after reforming. 10% oil (9.1 t o ll.S%), was introduced simultaneously at the
feed end of t h e trough, t o form a slurry of about 26 t o solids
by weight. Slurry temperatures did not exceed about YO" F.
Cooled Flakes A r e Soaked in Second Most The speed of t h e mixer conveyor was maintained at from 5 t o 8
Concentrated Miscella, Then Washed on Filter r.p.m.
Figure 7 is a flow diagram of t h e over-all process, including both The feed rate t o the filter for all runs was the equivalent in
the DreDaration
* - and the extraction Dhases. The filtration-ex- slurry form of 5 Dounds of cooked. cooled material Der minute.
traction phase is illustrated specifically in Figure 8. The speed of the %inch slurry feed.conveyor was regilated t o 50
250 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 45, No. 1

seed which had not been re-rolled. The residual lipidcs for run
5 , on flakes from subquality seed, mas l.570. These results
indicate t h a t for prime seed the slight gain in oil rccovery may
probably not justify the expense of the re-rolling step Re-
rolling of subquality seed would probably also not be justified,
since it is generally recognized that there is an economic level
for residual oil content depending on the particular cottonseed
and its grade, below which the increase in oil recovery would
be offset by the lowering of the quality of the oil with respect t o
the r e h i n g loss and refined and bleached color.
The residual lipides would be further reduced when hull
material is added t o the meal t o adjust the proteincontent to41%,
the standard for most commercially produced meals. The ex-
tent of the reduction would drprnd, of course, on the oil content
of the hull material. For thc meals produced in the runs re-
poited, the reduction could be significant, since the protein
contents (not tabulated) ranged between 48 and 51%.
Data in Table I11 show that thc major portion of the lipides
was removed during the initial blurry filtration and during the
first Rash-only 1.2 to 2.9% remaining in the cake after t h e
first wash. The low lipides and solvent contents of the cake
aftcr carh \\ash indicate the high efficitncy of drainage and wash-
ing that was obtainable by the UEC of vacuum. Thorough drain-
age following the final wash is of especial importance, since the
retained solvent contains some lipides.
The concentration of lipides in the first filtrate (final miscella)
averaged about 25% (Table 11),in comparison with 15 to 20%
usually obtained in final miscellas in conmercial direct extraction
operations. In the three filtrates used in the countercurrent
washing, the concentrations of lipides fall off markedly, as noted.
The higher concentration 111 the final miscella means that lesv
Figure 3. Top V i e w of Filter Showing M a r c Discharge Scroll solvent must be evaporated, thereby reducing steam consumption
as \vel1 as size of equipment required. With meats originally of
higher lipides content and where the hull rontcnt of the material
t o 60 r.p.m. t o dclivcr slurry to the filter pan at a rate equal t o feed stream can be reduced, the niiscella concentration with
the combined rate of the flakcs and the No. 2 filtrate dclivcred t o filtration-extraction could be increased.
the mixer. The filter was operated with a cakr thickricss of 1 AIiscclla concentrations as high as 28% have bcen achieved in
inch, pan speeds of 1.4 to 1.9 min. per rev., arid with vacuum re- other pilot-plant runs, with filtration-extraction, to be reported
quirements of 1.5 to 6.0 inches of mercury. The cake thickness elsewhere, where the solvent-to-flake ratio was only 1.0 t o 1.
was controlled by the speed of the pan. The solvent-to-flake These iesults indicated that there is considerable leeway for
ratio throughout was 1 1 to 1. efficimt operation and evtraction at still lower ratios with only a
The vacuum capacity employed ranged between 3 and 10 nominal sacrifice in exti action efficiencv and that further ad-
cubic feet per minute (volume a t the vacuum) per square foot vantages would hc realized hv the use of heated solvent. I n the
of filtering area. The blowback gas rate
ranged between 0.2 and 0.5 cubic feet per
minute per square foot of filtering area.
Periodic readings were taken of wash
rates, blowback gas rates, vacuum, and
cake thickness. Samples mere taken
periodically for analysis of the follow-
ing: feed material t o slurry mixer;
slurry; hexane; filtrates; filter cake be-
fore and after each wash; final discharged
marc: and h a 1 desolventized meal.

Extraction Results Compare Favor-

ably with Conventional Processes
Tables I1 and I11 give data on extrac-
tion; Table IV, data on filter capacity;
and Tables V and VI contain analytical
data indicating the quality of the oil and
meal products.
Residual lipides contents in the de-
solventized extracted cake (Table 11)
are in the desirable low range, averag-
ing 0.8% for runs 1 t o 3, on re-rolled,
cooked flakes from prime seed, compared
to 1.02% for run 4 on flakes from prime Figure 4. Slurry M i x i n g Conveyor at Left of Filter
January 1953 INDUSTRIAL A N D ENGINEERING CHEMISTRY 251

25% solvent requires the evaporation of only one

Table 111. Residual Lipides and Hexane Content of Cake on Filter ar third as much solvent per pound of solvent-free
Various Stages meal as a marc containing 50% solvent. More-
Run No. 1 2 3 4 5 over, any reduction in size of the marc desol-
Lip- Hex- Lip- Hex- Lip- Hex- Lip- Hex- Lip- Hex- ventizing unit is of particular significance because
ides, ane, ides, ane, ides, ane, ides, ane, ides, ane,
% % % % % % % % % %
this type of equipment is relatively large and ex-
Before 1st wash 11.7 28.3 10.3 25.2 12.0 2 8 . 1 14.7 30.9 13 2 26.7 pensive t o fabricate. Secondly, the lowcr marc
After 1st wash 2.131.0 1.2 25.6 1.526.4 1,936.1 2:b29.8 solvent content should result in a reduction in the
After 2nd wash 1 . 1 34.0 0.8 31.0 0.9 2 5 . 9 0 . 9 32.9 1 . 7 32.4
After 3rd wash 0 . 9 27.8 0.6 22.1 0.9 21.3 0 . 8 28.4 1 . 5 24.8 residual lipides of the final meal since solvent re-
tained in the marc contains lipides.
The solvent-to-flake ratio in filtration-extraction,
Table IV. Filter Capacities 1.1 t o 1 bv weiaht (Table II)-amreciablv less
(Basis 1250 lb. meats-hulls mixture per ton of cottonseed) than the ratio of about i.8 to-1 used in cottonseed processing
Tons Cottonseed per 24 Hours plants employing direct solvent extraction-is attributed t o the
Pilot plant filter Commercial filter"
Feed Rate. (3 ft. diam., (10ft. diam., high extractability of the cooked material, the efficiency of the
Lb./iMin. 3.5 sq. ft.) 65.0 sq. ft.) displacement washing, the effectiveness of the vacuum drainage
5 6 110 between washes, and the thorough drainage of the final marc.
10 12 220
15 18 335 The low solvent requirement coupled with the short slurry
20 24 445
holdup time for extraction permits smaller equipment and should
a Capacities calculated on basis of filtering area. decrease solvent losses as well as enhance the safety features of
the process.
While no particular effort has been directed so far to determine
Table V. Data on Crude Oils the minimum solvent-to-flake ratio obtainable t h a t would be
Crude Oil commensurate with practiral operation and efficient oil extrac-
Free Free Refining Lovibond Color
fatty gossypol, Loss", Refined Bleached tion, as this development work moves further along, i t should b e
Run S o . acid % % Y-R Y-R possible t o achieve much lower ratios through the uw of im-
1 0.74 0.013 2.1 36-5.8 20-1.2 proved preparation techniques, higher solvent temperatures, and
2 1.10 0.012 4.6 35-5.3 20-1.6
3 0.93 0.008 5.0 35-7.4 20-1.6 recircu1at:on and flooding of wash liquids on the filter Also,
4 0.97 0.009 4.2 35-5.3 15-1.7
5 5.23 0.089 15.6 35-8.3 35-2.7 for cottonseed, it should be pointcd out t h a t reduction in the hull
a A.O.C.S. official hydraulic refinihg method was used; all foots were firm. content of the meats stream may not necessarily make possible a.
lower solvent-to-flake ratio, but the solvent-to-whole Beed ratio
should bc lower.
Table VI. Data on Desolventized Meals Vacuum requirements as t o degrec and volumetric rate are in
Nitro en the desirable low operating range'and could have been reduced to
Run Lipides. Mois- GoSsYPol, % Solubfes, Thiamine, some extent in all the runs without adversely affecting operation
No. yo ture, % Free Total % P.P.M.
1 0.90 4.8 0 030 1.17 25 7
efficiency. Vacuum requirements will vary with the method of
2 0.66 5.1 0.033 1.33 38 1 15:6 material preparation, the amount of blowback gas applied, with
3 0.83 6.0 0.027 1.32 34.3 17 0
4 1.02 5 3 0.028 1.28 44.7 .. cake thickness, and with the type and size. of the filter screen.
5 1.49 5.1 0.050 1.25 41.3 .* In runs 4 and 5 , in which the material was not reformed and
therefore contained less fines, the vacuum required was only 3.0
and 1.5 inches of mercury, respectively. It should be pointed
out t h a t the vacuum can be varied as desired t o attain a condi-
batch-type investigations using unheated solvent, ratios as low as tion t h a t a n operator may consider advantageous. As a n ex-
0.85 t o 1 have been employed successfully, yielding miscella with ample, the extraction efficiency can be improved, if desirable, b y
as high as 32% oil content. decreasing the system vacuum, as this increases the pooling of
The fines content (0.13 t o 0.30% b y weight) in the first filtrates wash liquid on the cake.
(Table 11) is considered low for direct solvent extraction and is Rlowback requirements, as t o volumetric rate and pressure
especially low, considering t h a t the miscella is filtered through a (less than 1 pound per square inch) were also in the desirable low
cake of only 1-inch thickness. The fines form a much more range and will vary with the material characteristics, the degree
porous bed in a polishing filter than do fines from uncooked ma- of vacuum required, the type of screen, and size of screen open-
terial and can be filtered more readily. The concentration of ings.
fines in the second, third, and fourth filtrates averaged 0.07, Although nitrogen was used as the blowback gas in this in-
0.03, and 0.03y0 by weight, respectively. vestigation, i t is realized t h a t in translation t o commercial
It is probable t h a t the fines problem can be reduced consider- scale, i t would be more practical t o use the hexane-saturated air
ably or eliminated by permitting a longer slurry settling time for from the filter hood; however, gas from a noncombustible gas
formation of the original cake on the filter and by circulating the generator, concentrated miscella, or possibly superheated hexane
first filtrate miscella back t o the cake on the filter for refiltration- vapors could also be used.
a well-known technique used commercially-and draining it off
just ahead of the first wash. The wash sprays could be moved back Filter Capacity. Table IV gives filter capacities obtained with
the filter used in this investigation (3 feet in diameter), with the
one space and a fifth filtrate receiver installed, or the first wash corresponding capacities for a commercial filter (10 feet in diam-
eliminated. As shown in Table 111, the elimination of one wash eter), extrapolated on the basis of filtering area. Since these fig-
would leave only about 0.2% more residual lipides in the cake. ures are calculated on the basis of 1250 pounds of meah-hulls
Either method would be feasible; the one chosen would depend mixture fed t o the filter per ton of whole cottonseed, i t is obvious
that greater capacities than shown in the tabulation should result
mainly on economical considerations. where the particular seed and mills separating equipment will
Solvent content of marc t o the dryers ranged from 22 t o 28% permit better purification of meats, up t o a point of diminish-
(Table 11), as compared to the 45 t o 55% obtained commercially. ing returns as determined by the maximum allowance of oil loss
This lower solvent content is significant for two reasons. First, in the hull stream. T h e more hulls t h a t can be removed from the
meats stream, the less the material that must be handled by
i t should result in a n appreciable saving in steam consumption the rolls, cooker, slurry-mixer, filter, and desolventizers, and t h e
and in smaller desolventizing equipment, since a marc containing lower the requirement of solvent per ton of whole seed for extrac-
252 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 45, No. 1

The free gossypol contents in the crude oils are low, showing
t h a t the cooking procedure advocated herein for filtration-ex-
traction was adequate for chemically combining or binding the
bulk of the free gossypol in the meal. I t is common knowledge in
the industry that crude oils of low gossypol content usually ex-
hibit only slight color reversion on storage a t 100" F. for as long
as 30 days.
Additional refining data, not included in the table, were ob-
tained on hydraulically pressed crude oils prepared from separate
portions of the cooked flakes from each of the pilot plant filtra-
tion-extraction runs for comparison with the filtration-extractcd
oils. The latter and the corresponding hydraulic oils were
closely comparable in refining loss and in refined a.nd bleached
color, both as prepared and after 14 days of storage a t 100" F.
Table VI gives essential quality data pertaining t o the final
(desolvent'ized) meals. The values for free gossypol (0.027 to
0.033%) are appreciably lower than those normally obtained in
commercial hydraulic pressing operations with prime seed.
This lower free gossypol content is attributed to more effective
rolling and t o the use of higher moisture contents of the flakes
during cooking, The free gossypol content would be further
reduced by adding hull material to the meal t o adjust the protein
content to 41%. Nitrogen solubility and the thiamine contents
of the meals are higher than normally obtained in hydraulic
pressing. These results are attributable to the shorter cooking
time and to the lower cooking teniperaturo.

Figure 5. Filtrate Receiver Assembly

tor. The additional hulls removed would, of c o u r s ~he , addrd t o

the extracted meal for adjustment t o 41y0 protein content.
The maximum capacity a t M hich the pilot plant filter \?as oper-
ated was equivalent to a rate of 24 tons of cottonseed per 24-hour
day. The factor that limited the pilot plant capasit3 Kith prop-
erly prepared material was the actual physical handling of the
large liquid volumes required for the higher material feed rate on
such a small filter area. These higher rates would not neces-
sarily be recommended for commercial practice. For example,
for a 150-ton-per-day plant, the 10-foot filter might be desired in
place of, say, a 6-foot filter with a filtering area of 25 square feet
(calculated capacity 170 tons) since a smaller cake thickness Figure 6. O n 1 4 - M e s h Fraction of Typical C o o k e d Unreformed
could be used and the difference in cost of the two filters is rela- Material Showing Hulls Curled around M e a t s Particles
tively small. This would give leewav for operating above plant
capacity, if desired, or to obtain highest possible capacities with
oil-bearing materials having inferior extractability properties or These analytical values for any givcn cottonseed m e d arc
having lower filterability characteristics-for example, materials
that contain excessive fincs or that have not been properly pre- being correlated with rcsults in feeding experiments to establish
pared for extraction. whether such chemical tests can bo used t o indicate the nutritional
value of a meal for livestock.
Short Moist Cook at Low Tem-
peratures Improves Product Quality Cost Estimates for Commercial Installation
Indicate Advantageous Investment Rates
Analysee of the crude oil produced for free fatty acid and free
gossypol contents wi'h refining data are given in Table V. The interest manifested by the oilseed crushing industry when
Bleachable prime oils werc obtained from the prime seed, and the the details of the filtration-extraction process were first made avail-
oils produced from the subquality seed refined and bleached t o able indicated the need for a cost analysis as early as possible.
just beyond the upper limits for prime color of 7.5 red and 2.5 red, A preliminary cost study ( 1 5 ) of the application of the process
respectively. The low bleach colors are of particular iniportance t o cottonseed, based on the available pilot plant development and
since oil refiners consider bleach color the best single criterion material balance data on filtration-extraction and on the quota-
of cottonseed oil quality. tions of equipment manufacturers for conventional commercial
January 1953 INDUSTRIAL AND ENGINEERING CHEMISTRY 253

COTTONSEEO costs and also provide information necessary for detailed designs
to meet the requirements of particular situations.
LINTERS
At least two equipment fabricators are reportedly preparing
HULLS engineering evaluations of the process in contemplation of
bidding on commercial-sized installations.
FLAKES
Development Work Promises
Successful Application to Other Oilseeds
1 MILD COOKING 1
While the process development work has been mainly with
I CONDITIONING cottonseed, favorable results have been obtained on a limited
number of preliminary pilot plant runs on soybeans and on raw
I AGITATION WITH I and converted rice bran.
SOLVENT OR MISCELLA
I
Some other oil-bearing materials t h a t have been investigated
on a preliminary, batch scale with promising results include pea-
nuts, flaxseed, and tung press-cake. These small scale tests,
CONTiNUOUS HORIZONTAL
which have been specifically designed for extrapolation t o the
pilot plant scale, have correlated satisfactorily in the work done
so far with cottonseed, soybeans, and rice bran.
For processing oil-bearing materials other than cottonseed,
the indications are that optimal preparation and extraction con-
ditions will not be critical, but will be somewhat different from
one material to another. Versatility of the filtration-extraction
process for handling other oilseeds is of particular importance t o
cottonseed oil mills that supplement their seasonal operation by
processing soybeans, peanuts, and other oleaginous crops that are
becoming available in larger and larger volume in the Southern
area.
While the process as described operates satisfactorily under
Figure 7. Flow Diagram of Process for Cottonseed
the described operating conditions, there is, as with any process
in its early Rtage of technological development, latitude for
equipment units for the remainder of the plant (exclusive of
office buildings, control laboratory, and steam-generating plant),
indicates t h a t the over-all investment cost for a completely in-
stalled plant for processing 100 tons of cottonseed a day would
be somewhat lower than that of a comparable conventional direct
solvent extraction plant. Comparative cost data from private
d SLURRY MIXER-CONVEYOR k
sources (not given in the study) indicate t h a t the estimated in-
stalled cost for only the extraction and desolventization equip-
ment would be approximately 15% lower than for a conventional
direct extraction plant.
The cost of converting a direct solvent-extraction plant would
be the cost of replacing the extractor unit and accessories with the
slurry-mixer, filter, and accessories. The equipment costs, not
including installation costs for these, are given in the cost study.
On converting a n existing screw-press or hydraulic-press mill t o
filtration-extraction, the filtration-extraction equipment would
replace the pressing equipment. The estimated costs listed in
the cost study (16)under “Filtration-Extraction Unit” would
represent t h e approximate probable conversion costs for in-
stallations of 50,200, and 400 tons capacity.
NO.PWASH
-- <
Low Solvent Requirements
Should Reduce Operating Costs
- S O L V E N T FEED N 0 . 3 WASH

Figure 8. Flow Diagram of Filtration-Extraction Phase of

Detailed data are not yet available for estimating operating Process
costs. Lower utility costs are indicated by the fact t h a t com-
pared to conventional direct solvent extraction, only one third
as much solvent may have t o be removed by the meal desolventiz- simplification and improvement. A better understanding of the
ers, and only one half as much by the oil recovery equipment. principles and variables involved in each of the unit operations
Also, the absence of critical operating conditions, the indicated should result in the selection of optimum processing conditions
ease of control of the filtration-extraction operations, and the for both the preparation and the extraction phases t h a t will
resulting lower supervision and operating costs anticipated indi- utilize the simplest types and fewest number of equipment units.
cate t h a t considerable savings could be expected. These factors Progress made along these lines should result in higher extraction
should also ensure low down-time and greater uniformity of prod- efficiency, lower solvent requirement, improved operability,
ucts. increased capacity, improved quality of oil and meal products,
Experimental operation of a commercial direct solvent extrac- and lower cost of installation and operation.
tion plant which is being adapted t o filtration-extraction should The results obtained to date, in which a reasonably wide range
provide a &mer basis for estimating investment and operating of preparation conditions was employed, showed that satisfactory
254 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 45, No. 1

filtration ratw and extractability and smooth performance of the have many self-regulating features; the integrated p r o c ~ s sshould
filter at relatively high capacities and a t low vacuum and blow- require no more operational attention than other solvent proc-
esses
back requirements could be attained. However, meats condi- Minimum time required to place the procese on stream
tioning and rolling, and the interrelationship of time, temperature, High capacity for small filter units
and moisture content in cooking are being further studied t o de- Lower installed cost of plants in comparison with existing
termine the permissible operating limits for each and the optimal processes
values that will yield material having the best possible filtration These advantages, coupled with the indicated versatility of
and extraction characteristics and concurrently produce oil and the filtration-extraction technique for direct processing of oil-
meal products of higher quality. seeds other than cottonseed as well as oil cakes and other olcag-
To determine optimal preparation conditions for cottonseed, inous materials, should make the process particularly attractive
owing t o the variations in processing characteristics between one to small- and medium-size mills.
lot of cottonseed and another, it will be important in the de-
velopment work to cover an adequate range of cottonseed repre-
Acknowledgment
sentative of varieties, growth locations, grades, and storage
history, Knowledge gained in processing cottonseed will be The authors wish t o acknowledge the close cooperation and as-
very helpful in ascertaining optimal preparation procedures for sistance of the Mechanical Service Division of this Laboratorj
other oil-bearing materials. in the installation of the pilot plant equipment. They also ex-
I n the extraction phase for cottonseed, improvements should press appreciation to the Analytical Section of the Analytical and
be realized from a concerted study and evaluation of each of the Physical Division for some of the analytical determinations and
many interrelated variables involved. to J. E. Hawkins and L. L. Holzmthal for preparation of the How
I n the operations prior t o the filtration step, t h e most im- diagram drawings.
portant factors for investigation include: extraction and filtration
characteristics of the slurry as affected by preparation conditions Literature Cited
prior t o extraction; temperature, retention time, and solids con-
centration in slurrying; improvement in design and type of slurry (1) Bonotto, M., Oil & Soup, 23 (9), 297-9 (1946).
(2) Brownell, L. E., and Katz, D. L., Chem. Eng. P r o g r e ~ 43,
, 537-
mixing unit so as to achieve minimum comminution and short- 48 (1947).
circuiting. (3) Coats, H. B., and Karnofsky, G., J . Am. Oil Chemists’ SOC., 27
In the filtration step, there is latitude for further reduction of (2), 51-3 (1950).
residual lipides in meal, solvent-to-flake ratio, miscella fines COR- (4) Cornell, D.,and Katz, D. L., IND. ENG.CHEM.,43,992-6 (1951).
(5) Gardner, H. K., D’Aquin, E. L., Parker, J. S.,and Gastrock,
tent, vacuum requirement, and also for increasing miscella con- E. A., Ibid., 44, 2261-4 (1952).
centration. (6) Gastrock, E. A,, and D’Aquin, E. L., Oil Mill Gaz., 53 (4),13-21
(1948).
(7) Gastrock, E. A., D’Aquin, E. L., and Spadaro, J. J., Itiid., 57,
Process Offers Balance for No. 1, 26-9 (1952).
Present Seasonal Production (8) Gastrock, E. A., D’Aquin, E. L., and Vix, H. L. E., Cotton Gin.&
Oil Mill Press. 53. 31-7 ihlav 1952).
The pilot plant development of a new, continuous process for (9) Gastrock, E. S.; Eaves, P. H:, and D’Aquin, E. I,., 0 z E Mill
the solvent extraction of cottonseed uses a continuous, horizontal, Guz., 57, NO. 1, 62-3 (1952).
rotary vacuum filter. The experimental filter has a filtering (10) Grace, H. P., Chem. Eng. Progress, 47 (lo), 502-7 (1951).
(11) Hill. R. N.. and Kaiser, 31. hl., in -4m. Inst. Chem. Engrs.,
area of only 3.5 square feet and was operated a t capacities equiva- “Symposium on Relationship between Pilot-Scale and Com-
lent t o rates up t o 24 tons of whole cottonseed a day. mercial Chemical Engineering Equipment,” pp. 57--77,
The process is rather simple, departing from conventional White Sulphur Springs Neeting, lLlaroh 11-14, 1951.
methods of processing in incorporating a mild cooking operation (12) Karnofsky, G., J . Am. Oil Chemists’ Soc., 26 (lo), 5 7 0 4 (1949).
(13) Kenney, J. A., in Am. Inst. Chem. Engrs., “Symposium on
with direct solvent extrartion and in t h a t the actual oil extraction Relationship between Pilot-Scale and Commercial Chemical
phase consists of a simple operation of mixing miscella and cooked Engineering Equipment,” pp. 46-56, White Sulphur Springe
flakes followed by filtration and countercurrent washing. The Meeting, March 11-14, 1951.
method of material preparation is such that it promotes oil ex- (14) Moore, N. H., Oil Mill Gaz., 54 (lo), 13-16 (1950).
(15) Persell, R. M., Pollard, E. F., Deckbar, F. A., Jr., and Gastrock,
traction and facilitates the washing action on the filter and is also E. A., Cotton Gin & Oil M i l l Press, 53 (17), 18-20 (1952).
responsible for producing oil and meal products of high quality. (16) Persell, R. hl., Pollard, E. F., and Gastrock, E. A., Ibicl., 51
The advantages of the process are as follows: (26), 9-11, 24-5, 30-2 (1950).
(17)
. . Reuther. C. G.. Jr., LeBlanc, M. F. H., Jr., Batson, D. LT.,and
Extraction t o a residual lipides of 1% or less Knoopfler, N. B., J. Am. Oil Chemists’ Soc., in press.
Low fines content in miscella (18) Spadaro, J. J., Graci, A. V., Gardner, H. K., Parker, J. S.,
Lower solvent requirements, which should decrease solvent Laborde, E. J., and Gastrock, E. A, Oil Mill Gaz., 56 ( l ) ,
losses and increase safety 77-81 (1951).
Lower solvent content of extracted cake and of final miscella
which should result in lower equipment and steam costa for sol- RECEIVED for review September 2, 1952. ACCEPTED October 30, 1982.
vent recovery Presented a t the American Oil Chemists’ Society meeting, Houston, Tcx.,
Conventional hydraulic preparation operations can be used April 28-30, 1952. The Southern Regional Research Laboratory iY one of
without expensive modifications the laboratories of the Bureau of Agricultural and Industrial Chemistry,
The filter is a slowmoving device, and it and its accessories Agricultural Research Administration, V, S. Department of Agriculture.