Improvements to laboratory-scale maize wet-millingprocedures

Michael K. Dowd *

Commodity Utilization Research Unit, Southern Regional Research Center, ARS, USDA, 1100 Robert E. Lee Blvd., New Orleans, LA

70124, USA

Received 15 July 2002; accepted 24 February 2003

Abstract

The wet milling of maize is difficult to study in the laboratory because some of the required separation steps are

challenging to implement at bench-scale. This work was conducted to develop an improved 100-g wet-milling procedure

that better models the industrial process. Several separation steps were modified from previously reported methods.

Among the changes, germ was recovered by a flotation/skimming technique that is normally used on larger-scale

procedures. Starch was recovered by tabling, but the flow profile at the end of the table was changed to reduce gluten

settling and the partitioning and pumping of slurry fractions was changed to allow the tabling process to begin

immediately after fiber recovery. Gluten was dewatering directly on the table overflow, and starch was recovered from

the table before drying. These modifications eliminated some problems associated with other procedures, e.g. the

scraping of tabled starch to reduce protein contamination, the loss of germ due to size reduction, and the separate

recovery of coarse and fine fiber fractions. Compared with routine tabling methods, the modified method used in this

work produced starch with less protein (0.42 versus 0.55% for the maize variety tested); however, the improvement was

achieved at the expense of a slightly lower starch yield (64.4 versus 65.4%). Standard deviations for the product yields

were 0.28% for starch, 0.27% for gluten, 0.24% for fiber, 0.13% for germ, and 0.07% for total solubles. The procedure

will be beneficial for some maize wet-milling experiments.

Published by Elsevier Science B.V.

Keywords: Corn; Maize; Milling; Processing; Separations; Wet milling

1. Introduction

Over the past 50 years, several laboratory and

pilot-plant wet-milling procedures have been de-

veloped to recover starch from maize. These

procedures were designed to study the industrial

process and to recover products similar to those

produced in industry. Because of the large scale of

the industrial process, even small differences in

yield or composition can have a significant impact

on milling profitability. Laboratory-scale proce-

dures, therefore, must be precise to be able to

detect the small yield or compositional differences

that may be commercially important. Conse-

quently, the precision of these milling techniques

is of continuing interest.* Tel.: �/1-504-286-4339; fax: �/1-504-286-4367.

E-mail address: mkdowd@srrc.ars.usda.gov (M.K. Dowd).

Industrial Crops and Products 18 (2003) 67�/76

www.elsevier.com/locate/indcrop

0926-6690/03/$ - see front matter. Published by Elsevier Science B.V.

doi:10.1016/S0926-6690(03)00034-7

In general, maize wet milling includes steps forsteeping, milling to free germ, germ separation,

milling to free starch, fiber recovery, and separa-

tion of starch and gluten. Many variations of these

steps have been reported, and Singh and Eckhoff

(1996) have reviewed much of the early work.

Although larger-scale procedures are still used,

two 100-g procedures developed by Eckhoff et al.

(1996) and Singh et al. (1997) form the basis ofmost current bench-scale experiments.

Several of the separation steps associated with

the industrial wet milling are difficult to implement

in the laboratory. As a consequence, the basis of

the separation methods used in laboratory proce-

dures is not always true to the industrial process.

To improve the similarity of laboratory and

industrial wet milling, several separation stepsfrom prior methods were modified. Changes were

made to better model the basis of the industrial

separation, to yield products that better reflect

industrial products, and to improve measurement

precision. The resulting procedure eliminates some

difficulties associated with other laboratory pro-

cedures and is a better model of the industrial

process.

2. Materials and analytical methods

Results reported in this work are for a commer-

cial high-protein yellow-dent maize variety. The

maize was dried to �/10% moisture in ambient air

and was stored at 4 8C. Samples were hand-picked

prior to milling to remove broken kernels, piecesof hulls, and foreign material.

Kernel moisture was determined according to

AACC Method 44-15a (2000). Starch was mea-

sured for the whole kernels and milled fiber

samples by polarimetry (CRA Method A-20,

1980). Nitrogen was measured for the kernels

and the milled starch and gluten samples by

combustion with a LECO (St. Joseph, MI, USA)FP-2000 nitrogen analyzer (AACC Method 46-30,

2000). A conversion factor of 6.25 was used to

convert nitrogen to protein. All recovered pro-

ducts were dried by a two-step oven procedure

based on AACC Methods 44-15a and 44-19

(2000). The sodium metabisulfite and lactic acid

used in steeping were from J.T. Baker (Phillips-burg, NJ, USA).

Tests for statistical differences were made with

the SAS GLM procedure for analysis of variance

(SAS Institute Inc., 1999). Pairwise t-tests were

calculated with the MEANS function and a�/

0.05.

3. Description of the procedure

Several series of experiments were conducted

during the development work, which led to a finalprocedure that recovered steepwater, germ, fiber,

starch, gluten, starch filtrate solubles, and gluten

filtrate solubles (Fig. 1). Two parallel milling

systems were built and were used concurrently on

several projects. During the initial studies with

these systems, a single maize variety was milled

before and after individual experiments as a

control. Because maize can be stored for longperiods without significant changes to laboratory

milling results (Singh et al., 1998), these control

runs were used to test the constancy of the

procedure. Analysis of these runs indicated that

no significant changes occurred within the systems

over the duration of the experiments (�/1 year)

and that no significant differences existed between

the two systems. Therefore, the data from thecontrol experiments were pooled and used to

estimate the precision of the procedure (Table 1).

3.1. Steeping (Stage A)

For steeping, 200-ml of a solution containing

0.5% (w/v) lactic acid and 0.2% (w/v) sulfur

dioxide (added as sodium metabisulfite) was added

to 100 g of kernels in a 500-ml screw-top flask.

Steeping was conducted for 24 h at 529/18C in an

orbital water bath operated at 175 rpm (Labline

Instruments, Melrose Place, IL, USA). Aftersteeping, the kernels and steepwater were sepa-

rated on a screen mounted in a funnel, and 50-ml

of water was used to rinse the steeping flask,

kernels, and funnel. The steepwater and rinse

volumes were combined in preweighed aluminum

pans and dried.

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/7668

3.2. First (coarse) grind (Stage B)

A 1.25-l Waring blender flask modified with

blunt-edged blades was used for the first milling

step. The blades were constructed from 6-mmaluminum bar stock and were similar in shape to

the blades described by Eckhoff et al. (1996). After

transferring the kernels to the blender flask, 200-

ml of fresh water was rinsed through the steeping

flask and funnel. This rinse water was added to the

kernels, and the mixture was blended first at 4700

rpm for 3 min and then at 5200 rpm for 3 min.

3.3. Germ recovery (Stages C and D)

Germ was separated from the first-grind slurry

by a flotation/skimming technique based on den-sity difference. The recovery of germ by this

technique is facilitated by having a slurry depth

sufficient to resolve the kernel components, but

with a large surface area to collect the germ. At the

100-g scale, this was best achieved with a 145-mm

diameter plastic beaker that was cut to a 180-mm

depth to allow better access to the slurry surface.

After transferring the slurry into the beaker, 50 ml

of clean water was rinsed through the blender flask

and was added to the slurry. This water addition

yielded slurry with a bulk density suitable for

skimming.

Briefly stirring the slurry temporarily suspends

the starch particles, which causes the germ and

coarse fiber particles to rise to the slurry’s surface.

The fiber settles at a slightly faster rate, leaving the

germ concentrated at the surface for a brief period

before it also begins to settle. In larger-scale

skimming procedures, the presence of fiber at the

surface reduces the time available to collect germ.

In this work, forceps were used to recover germ

individually during the first 90 s of the cycle, after

which a screen was used to recover germ in bulk

for an additional 2 min and 30 s. The slurry was

then remixed, and the process was repeated.

Because the slurry contains only a small number

of germ (300�/400 whole germ), the use of forceps

significantly increased the amount of germ recov-

ered per cycle and reduced the number of cycles

Fig. 1. Schematic diagram of the 100-g wet milling procedure. The entry of wash water into the process is depicted by dashed lines.

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/76 69

needed to degerm the slurry. Preliminary testing

was conducted to determine the optimal number

of mixing cycles needed to recover the germ.

During this testing, the seventh cycle increased

the germ yield by 0.049/0.01%. Because this

increase was less than the standard deviation of

the germ yield obtained with six mixing cycles

(Table 1), six cycles were used in the standard

procedure. Recovered germ was collected on a 7.6-

cm diameter 40-mesh sieve, and the entire opera-

tion was conducted over a tray to enable the

recovery of any lost drops of slurry.

Degermed slurry was transferred to the same

blender flask used for the first grind operation.

Germ was washed free of slurry solids on the 40-

mesh sieve with 125 ml of water. During this step,

the sieve was placed over the skimming beaker to

enable the germ wash water to be used as the first

wash of the skimming beaker and implements. The

beaker, implements, and tray were then washed

with a second 125-ml volume of water. Both

washes were added to the blender flask with the

degermed slurry.

3.4. Second (fine) grind (Stage E)

The second-grind operation was conducted in

the same blunt-bladed blender flask used for thefirst grind, but with the blender operated at 10,000

rpm for 3 min. The concentration of residual

starch in the fiber fraction (after washing) was

measured to confirm the effectiveness of the

technique (Table 1).

3.5. Fiber separation and washing (Stages F and

G)

Fiber was recovered on a 20.3-cm diameter 200-

mesh sieve that was fit onto the top of a 20.3-cm

diameter�/20.3-cm tall bucket. The assembly was

Table 1

Product yield and composition from a 100-g maize wet milling procedure and comparison of the repeatability of the procedure

(coefficients of variance) with other 100-g milling procedures

Minimum Maximum Mean Standard

deviation

CV CVa CVb

Milling yields (%)

Steepwater solids 4.09 4.32 4.21 0.07 1.7 2.3 2.2

Germ 5.56 5.97 5.75 0.13 2.3 5.8 12.8

Fiber 11.22 12.09 11.68 0.24 2.0 4.9 6.2

Starch 63.62 64.61 64.18 0.28 0.4 0.6 0.7

Starch filtrate solids 0.04 0.08 0.06 0.01 16.7 �/ �/

Gluten 9.71 10.52 10.05 0.27 2.7 3.4 6.1

Gluten filtrate solids 2.59 2.86 2.72 0.09 3.3 4.0 5.1

Total recovery (%) 98.49 99.18 98.65 0.19 0.2 0.4 �/

Total solubles (%) 6.89 7.08 6.99 0.07 1.0 �/ 2.2

Product composition (% db)

Protein in starch fraction 0.38 0.46 0.41 0.02 4.9 �/ 12.7

Protein in gluten fraction 47.6 50.2 48.6 0.8 1.7 �/ 5.1

Starch in fiber fraction 21.9 22.9 22.2 0.5 2.2 �/ �/

Starch recoveryc (%) 89.4 90.8 90.2 0.4 0.4 �/ 0.6

Protein recoveryd (%) 42.2 44.6 43.4 0.7 1.6 �/ �/

Procedure replicated 12 times for a control hybrid over a 1-year period.a From Eckhoff et al. (1996), n�/5.b From Singh et al. (1997). Values are averaged for three yellow dent corn hybrids (n�/3 for each hybrid).c Starch recovery�/starch yield�/100/kernel starch (db). Kernel starch (db)�/71.29/0.3%.d Protein recovery (as gluten)�/gluten yield�/gluten protein concentration/kernel protein (db). Kernel protein (db)�/11.249/

0.01%.

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/7670

mounted on a portable sieve shaker (Model RX-86, W.S. Taylor, Mentor, OH, USA), and with the

shaker operating, the second-grind slurry was

poured slowly over the sieve. The blender flask

was then washed with 500 ml of water, which was

also poured slowly onto the shaking sieve. The

shaker was operated for 5 min to allow all of the

slurry liquid to collect in the bucket.

For fiber washing, the sieve was transferred intothe bottom of a Rubbermaid Model 2963 9.5-l

bucket that was cut to a height just greater than

the sieve height (�/70 mm) to enable the bucket/

sieve assembly to be mounted on the sieve shaker.

Other buckets can be used for this step, but this

bucket has a raised bottom that centers the sieve

and reduces the amount of water needed for the

washing operation. To secure the bucket/sieveassembly to the shaker, a 20.3-cm diameter�/19-

mm thick wood plate with a circular slot cut into

its face (1.6-mm deep�/3.2-mm wide) was fitted

onto the shaker bed. The slot was sized to just fit

the bucket skirting and allowed the assembly to be

leveled and clamped to the shaker. Two hundred

and forty milliliters of water was added to the top

of the sieve, which was just sufficient to allow thefiber to move with resistance across the screen’s

surface. The shaker was operated for 5 min, after

which the sieve was carefully transferred to a

second bucket, and the process was repeated.

Most of the sub-200 mesh particles were recovered

with three successive 240-ml washes. During test-

ing, a fourth wash recovered an additional 0.139/

0.01% of the kernel dry mass. Because thisadditional mass was less than the variability of

the combined starch and gluten yields after three

washes, three washes were used in the standard

procedure. A final 125-ml of water was then used

to rinse the sieve-frame and fiber. The fiber was

then transferred into a preweighed aluminum pan

and dried.

3.6. Starch�/gluten separation (starch tabling)

(Stage H)

Most laboratory-scale milling procedures sepa-

rate starch from gluten by pumping the slurry onto

the raised end of an inclined open channel (starch

table) and allowing the denser starch particles to

settle as the slurry flows along the channel. In thiswork, a similar process was used but with mod-

ifications that included changes to the slurry-

pumping schedule, the table length, and the

inclusion of a 25-mm diameter stir-rod at the end

of the table to modify the slurry flow as it exits the

table.

In the modified pumping scheme, the initial

starch�/gluten slurry and fiber washes werepumped onto the starch table in four portions.

The concentrated slurry from the initial fiber

screening was pumped onto the table first. This

was followed by the combined first two fiber

washes (240 ml each), as a first table wash. The

third fiber wash (240 ml) and the final fiber rinse

(125 ml) were then combined and used as a second

table wash. Finally, 125 ml of clean water waspumped onto the table. This pumping schedule

enabled the tabling operation to begin immediately

after separation of the fiber.

Most 100-g procedures use starch tables con-

structed from aluminum U-channels (5-cm wide�/

2.5-cm deep�/244-cm long) inclined 0.0156 cm/cm

with a 50 ml/min slurry flowrate (Eckhoff et al.,

1996; Singh et al., 1997). A similar table geometrywas used in this work, except that a stir-rod was

extended from the gluten receiver and propped

against the center of the end of the U-channel at

an angle of 509/28 from vertical (Fig. 2). The

purpose of the rod was to reduce surface tension

effects that influence the velocity field at the end of

the table. Preliminary trials indicated that the stir-

rod significantly reduced the amount of proteindeposited at the end of the table (Table 2), but that

it also reduced starch yield. Consequently, tests

were conducted with 244- and 305-cm tables

(Table 2). These studies indicated that at either

table length, the addition of the stir-rod at the end

of the table reduced the starch yield and increased

the gluten yield. The concentration of protein in

both fractions was reduced, indicating that theflow change resulted in both starch and gluten

particles being carried over into the gluten slurry.

Increasing the table length with the stir-rod

increased the starch yield and increased the protein

concentration of the gluten fraction. A compar-

ison of the 244-cm table without the stir-rod with

the 305-cm table with the stir-rod indicated that

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/76 71

the former conditions produced a slightly greater

starch yield (Table 2) but that the latter conditionsproduced a starch product with significantly less

protein. The 305-cm table with the stir-rod was

adopted for routine work.

3.7. Gluten dewatering (Stage I)

Gluten in the table overflow was recovered

during the tabling by using a 180-mm vacuum

filter with preweighed 150-mm Whatman #54

filter paper as the gluten receiver (Fig. 2). Har-

dened filter paper was used for this operation, as

holes occasionally formed with nonhardened pa-

pers. In addition, the filter paper was periodically

wetted during the early part of the tabling to

prevent dry spots from forming that occasionally

allowed the gluten to by-pass the filter. For routine

steeping conditions, the dewatering process was

completed within a couple of minutes of the

tabling operation. For some nonroutine steeping

conditions (e.g. if steeping is conducted without

lactic acid), filtration times can be much longer,

and the filter flask must be set aside to complete

this step.The dewatered gluten and filter paper were

transferred to an aluminum pan and dried. The

yield of solubles in the gluten filtrate was estimated

by measuring the volume of the filtrate (�/2 l) and

drying a 100-ml aliquot. A calibrated measuring

Fig. 2. Schematic diagram of the tabling and gluten dewatering steps.

Table 2

Comparison of tabling conditions on the yield and composition of the starch and gluten fractions

Table length (cm) Stir-rod table modificationa Starch yield (%) Gluten yield (%) Protein in starch fraction (%) Protein in gluten fraction (%)

244 No 65.49/0.3 a 8.69/0.2 a 0.559/0.04 a 53.59/0.8 a

244 Yes 63.49/0.3 b 10.89/0.1 b 0.379/0.03 b 45.69/0.4 b

305 No 65.79/0.2 a 8.39/0.1 c 0.529/0.04 a 55.39/0.5 c

305 Yes 64.49/0.2 c 9.89/0.1 d 0.429/0.02 b 49.09/1.0 d

Values within a column with the same letter are not significantly different at P �/0.05 (n�/3).a The fluid flow path was modified by providing a surface to channel the fluid from the end of the table into the gluten receiver.

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/7672

cylinder and volumetric pipette were used for thisdetermination.

3.8. Starch dewatering (Stage J)

Starch was recovered from the table prior to

drying. For this step, the table was inclined 8�/108,and the starch was gently lifted and pushed down

the table with a plastic putty knife and a stream ofwater. The starch was dewatered as it was collected

on a second 180-mm vacuum filter flask with

preweighed Whatman #50 filter paper. The starch

and filter paper were then transferred to an

aluminum pan and dried.

Initial tests indicated that the filtrate from the

dewatering step contained a small amount of

solids (Table 1). This fraction was accounted forby measuring the filtrate volume and drying a 200-

ml aliquot. The accurate measurement of this

fraction was facilitated by keeping the volume of

water used to collect the starch to a minimum

(typically �/600 ml).

4. Results

The maize sample used in the work contained

71.29/0.3% starch and 11.249/0.01% protein.

Milling yields (Table 1) were similar to those

obtained from most laboratory-based procedures

(Singh and Eckhoff, 1996). The coefficients of

variance for the yields (Table 1) were slightly

smaller for this procedure compared with other

100-g procedures (Eckhoff et al., 1996; Singh et al.,1997); however, part of this difference may reflect

the different numbers of replicates represented in

the studies. Yields of total solubles, germ, and

fiber were similar to industrial values. The yield of

gluten was higher (and correspondingly the yield

of starch was slightly lower) from this procedure

than from industrial operations. This difference is

characteristic of milling procedures that use starchtables. In this instance, part of the difference may

also have resulted because of the high-protein

variety used in the work.

Total recovery of solids from this procedure was

98.69/0.2%. Recoveries of �/98% are typical for

laboratory procedures (Singh and Eckhoff, 1996).

The reasons for the less than complete recovery ofsolids are difficult to ascertain, particularly as the

weight of the steeping chemicals is usually ne-

glected in these calculations. However, part of this

difference is likely related to the different oven

methods used to determine kernel moisture and to

dry the wet-milled fractions.

The protein concentration of the starch was

0.419/0.02%, slightly greater than the industrystandard of 0.30�/0.35% (May, 1987). This was

also likely due to the high protein concentration of

the variety, as the procedure has yielded starch

with a protein concentration below 0.35% for

many maize varieties. The protein concentration

of the gluten was 48.69/0.8%, which is low for

typical industry samples of about 66%. This

difference is another characteristic of millingprocedures that use starch tables. The concentra-

tion of starch in the fiber was 22.29/0.5% (Table 1)

in reasonable agreement with industrial values of

about 20% (May, 1987). The texture of the fiber

was similar to industrial ‘white fiber’ and was

considerably different from the coarse fiber (re-

covered from germ) and fine fiber fractions

produced from other 100-g laboratory procedures.

5. Comparison of separation techniques with other

100-g milling procedures

5.1. Steeping

Laboratory steeping systems vary in complexity

from static batch systems to counter-currentcirculating systems. At the 100-g scale, steeping is

usually conducted in batch mode with a stagnant

water bath (Eckhoff et al., 1996; Singh et al.,

1997). Although most of the resistance to the

movement of chemicals and kernel components

during steeping is internal, the nearly continuous

pumping of steepwater that occurs in the industrial

process minimizes any external transport effects.Consequently, an orbital water bath was used in

this work to better represent this situation. For the

control hybrid, agitating the steeps at 175 rpm

increased the yield of steepwater solubles by about

0.15% and increased the yield of total solubles by

about 0.08%.

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/76 73

An analysis of error propagation (Young, 1962)indicated that for laboratory-scale milling proce-

dures volume errors are more likely to contribute

to yield variability than mass errors. Conse-

quently, efforts were made to avoid or improve

volume measurements. Combining the steepwater

and rinse volumes and drying the entire volume

eliminated any error associated with measuring

steepwater aliquots and volumes. However, ifsteepwater is needed for additional testing, then

the yield of steepwater solids can be estimated by

measuring the steepwater volume and drying an

aliquot, as is standard practice in other procedures

(Eckhoff et al., 1996; Singh et al., 1997). A

calibrated volumetric cylinder and pipette are

recommended for these measurements.

5.2. First grind

The first-grind step was not altered significantly

from previous methods. However, construction of

several of the flask/blade assemblies used in this

step produced starch yields that varied over a

range of 1.5%, which would likely produce greater

yield variability than is typically reported for theseprocedures. To minimize this variability, all runs

within an experiment are conducted with either a

single flask or a set of flasks that were initially

tested and found to produce similar yield and

compositional results. The data presented in this

work were obtained with a pair of flasks developed

in this manner.

5.3. Germ recovery

Germ recovery has always been difficult to

mimic in laboratory procedures. In industry, this

separation is achieved with germ-clones, a hydro-

cyclone technique that separates particles by

density. Although hydrocyclones have been used

at pilot-plant scale (Rubens, 1990), the technique

is difficult to implement on small batches of maize.Large-scale laboratory procedures (]/300 g) typi-

cally recover germ by a flotation/skimming process

(Eckhoff et al., 1993; Singh et al., 1997), which is

also based on density difference. Often the time

allotted for skimming is to improve the repeat-

ability of the technique; still, the process is difficult

to standardize. At the 100-g scale, a sieving step isoften used to recover germ and coarse fiber

together, and the coarse fiber is separated by

aspiration after the combined sample is dried

(Eckhoff et al., 1996; Singh et al., 1997). The

advantage of sieving is that it is easy to standar-

dize. The disadvantage of sieving is that the

separation is based on particle size rather than

density difference. Sieving to recover germ isespecially problematic when a significant degree

of germ damage has occurred during the first-

grind procedure.

The flotation/skimming process used in this

procedure was very reproducible. Beaker shape,

the number and duration of the mixing cycles, and

the timed use of forceps and screens were all fixed

to ‘standardize’ the process as much as possible.With this procedure, a single operator was able to

produce germ yields with a coefficient of variance

of 2�/3% (Table 1). Because the coefficient of

variance was higher when samples were replicated

by more than one operator (typically 5�/7%), a

single operator routinely degerms all of the

samples within a given experimental set. Regard-

less of the number of operators, the consistency ofthe method compares favorably to the sieving

techniques used in other 100-g procedures (Table

1).

Because germ skimming is a ‘hands-on’ process

that is prone to small losses of slurry and these

losses are more critical at the 100-g scale, the entire

process was conducted over a tray. The tray was

then included in the germ washing operation. Theloss of solids at this step appeared to be insignif-

icant, as the total recovery of solids from this

procedure was comparable to other 100-g proce-

dures (Table 1).

In the standard procedure, 50 ml of water is

used to dilute the slurry to a density suitable for

germ flotation. For an occasional difficult hybrid

or for less effective steeping conditions, adding lesswater improved germ flotation. If necessary, the

bulk density of the first-grind slurry can be

measured before the water addition, and the

volume adjusted accordingly.

Overall, the advantages of skimming at the 100-

g scale are that the separation is based on density

difference and the separation of coarse fiber from

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/7674

germ is eliminated. The hand recovery of brokengerm pieces is also eliminated. The disadvantages

of the technique are that each sample requires 25

min to degerm and the process is difficult to

implement on a high-throughput basis.

5.4. Second grind

At the 100-g scale, second-grind operations have

been conducted with either blender flasks operated

at high speed (Singh et al., 1997) or attrition disk

mills (Eckhoff et al., 1996). In this work, the same

blender flask used for the first-grind step was usedfor the second-grind step, but with the blender

operated at high speed. The low concentration of

starch left with the fiber fraction (Table 1)

indicated that the approach was suitable. The

benefit of using the same flask for both milling

steps was that it reduced the number of vessels and

the amount of wash water used in the procedure.

One potential difficulty was found with thisstep. The standard 1.25-l Waring blender flask is

just large enough to hold the slurry and all of the

specified wash water accumulated through the

germ recovery step. If any additional water is

used prior to the second grind, the total slurry

volume can exceed the operating capacity of the

blender flask. This situation forces the use of either

a larger flask or a two-stage second-grind proce-dure.

5.5. Starch�/gluten separation

Starch tabling is problematic in that some gluten

tend to deposit at the end of the table. In some

procedures, a stream of water is used to gently

push the gluten particles off the table (Eckhoff et

al., 1993). In other procedures, a small portion of

the most contaminated starch at the end of the

table is scraped into the gluten fraction (Singh et

al., 1997). These practices reduce the amount ofprotein in the recovered starch product, but also

reduce the protein concentration of the gluten and

increase the variability of the starch and gluten

yields. If yield reproducibility is an important

component of the experiment, the settled gluten

is often accepted as part of the starch product.

Observation of the tabling process suggestedthat the gluten build-up occurred because of a

slowing of the slurry velocity near the end of the

table. As the fluid separates or ‘breaks’ from the

table, surface tension tends to raise the fluid depth.

At a constant volumetric flow rate, the thickening

lowers the average local fluid velocity, which

hastens the settling of gluten particles. The stir-

rod used in this work provided a path for theslurry to flow from the table, which eliminated the

surface tension effects and reduced the accumula-

tion of gluten particles at the end of the table.

The sequence for pumping the various starch�/

gluten slurry fractions onto the table was also

modified in this procedure. In other procedures

(Shandera et al., 1995; Eckhoff et al., 1996; Singh

et al., 1997), the fiber separation and washingssteps result in a single starch�/gluten slurry. This

slurry is allowed to settle for 0.5�/1 h (overnight, in

some cases) before it is decanted to give a

concentrated slurry fraction and a diluted slurry

fraction. Typically, the concentrated fraction is

pumped onto the table first, followed, in succes-

sion, by the diluted fraction and a water wash. By

making small modifications to the fiber separationand washing steps, the initial starch�/gluten filtrate

and the various fiber washes were maintained

separate. This eliminated the decanting process

and enabled starch tabling to begin immediately

after fiber recovery.

By modifying the table flow and length, some

fine-tuning of the yields and compositions of the

starch and gluten fractions is possible (Table 2). Ifdesired, multiple runs at different tabling condi-

tions can be used to recover starch and gluten

products with compositional properties that ap-

proach those produced in industry. Even with

these manipulations, however, the one-step settling

of starch on an open channel is not able to achieve

the sharp starch�/gluten separation of the contin-

uous multiple-stage industrial process.

5.6. Gluten dewatering

The standard process for dewatering gluten is

time-consuming (Eckhoff et al., 1996), and often

the gluten yield is estimated by drying an aliquot

of the gluten slurry. Because this aliquot contains

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/76 75

soluble components initially present within thekernel, the practice results in a high yield of gluten

and low yield of total solubles. By dewatering the

gluten as it flows from the table, the gluten

particles accumulate on the filter gradually over

the course of the tabling operation, and the flux of

water through the filter is increased for most of the

process. Under normal milling conditions, the

gluten dewatering step is completed at essentiallythe same time as the tabling operation.

5.7. Starch dewatering

Finally, in other 100-g procedures (Eckhoff et

al., 1996; Singh et al., 1997), the starch is initially

dried on the surface of the table. The table and

starch are then weighed together, and the starchyield is determined by difference. In this proce-

dure, the starch was recovered wet. The advan-

tages of this change were that it eliminated the

need to maintain accurate weights for the tables

and it reduced the number of tables used during

daily operation. The time required to recover the

starch wet (15�/20 min) was approximately equal

to the time required to recover the dry starch andclean the table.

6. Summary

An improved 100-g maize wet-milling procedure

is described with modified germ separation, fiber

washing, starch�/gluten tabling, and gluten dewa-

tering steps. The method recovers steepwater,starch, germ, gluten, fiber, and gluten and starch

filtrate fractions. Total solids recovery and yield

reproducibility was comparable to the best of

previously reported procedures (Table 1), and the

process was very stable over a 1-year test period.

The basic procedure, from the end of the steeping

step until the beginning of oven drying, requires

about 3.25 h, and the method can be implementedon a daily morning and afternoon schedule. Germ

recovery requires more time in this procedure, but

the time lost is compensated for by the elimination

of the starch�/gluten decanting step and the

reduced time associated with gluten dewatering.

Some problems associated with other procedures

are eliminated, such as germ loss due to sizereduction, the separation of fine and coarse fiber

fractions, and the scrapping of table ends to reduce

the protein concentration of the starch product.

Acknowledgements

The author thanks Scott Pelitire, Matthew Rice,

and Gasper Migliore, III for help with the

laboratory work.

References

American Association of Cereal Chemists, 2000. Approved

Methods of the American Association of Cereal Chemists,

10th ed. American Association of Cereal Chemists, St. Paul,

MN, USA, Methods 44-15a, 44-19, and 46-30.

Corn Refiners Association, 1980. Standard Analytical Methods

of the Member Companies, 6th ed. Corn Refiners Associa-

tion, Inc., Washington, DC, USA, Method A-20.

Eckhoff, S.R., Rausch, K.D., Fox, E.J., Tso, C.C., Wu, X.,

Pan, Z., Buriak, P., 1993. A laboratory wet-milling proce-

dure to increase reproducibility and accuracy of product

yields. Cereal Chem. 70, 723�/727.

Eckhoff, S.R., Singh, S.K., Zehr, B.E., Rausch, K.D., Fox,

E.J., Mistry, A.K., Haken, A.E., Niu, Y.X., Zou, S.H.,

Buriak, P., Tumbleson, M.E., Keeling, P.L., 1996. A 100-g

laboratory corn wet milling procedure. Cereal Chem. 73,

54�/57.

May, J.B., 1987. Wet milling: process and products. In:

Watson, S.A., Ramstad, P.E. (Eds.), Corn: Chemistry and

Technology. American Association of Cereal Chemists, St.

Paul, MN, USA, pp. 377�/397.

Rubens, R.W., 1990. A pilot plant for the wet milling of corn

grain. Cereal Foods World 35, 1166�/1169.

SAS Institute Inc., 1999. SAS/STAT User’s Guide, Version 8.

SAS Institute Inc., Cary, NC, USA.

Shandera, D.L., Parkhurst, A.M., Jackson, D.S., 1995. Inter-

actions of sulfur dioxide, lactic acid, and temperature during

simulated corn wet milling. Cereal Chem. 72, 371�/378.

Singh, N., Eckhoff, S.R., 1996. Wet milling of corn*/a review

of laboratory-scale and pilot plant-scale procedures. Cereal

Chem. 73, 659�/667.

Singh, S.K., Johnson, L.A., Pollak, L.M., Fox, S.R., Bailey,

T.B., 1997. Comparison of laboratory and pilot-plant corn

wet-milling procedures. Cereal Chem. 74, 40�/48.

Singh, V., Haken, A.E., Niu, Y.X., Zou, S.H., Eckhoff, S.R.,

1998. Wet-milling characteristics of selected yellow dent

corn hybrids as influenced by storage conditions. Cereal

Chem. 75 (2), 235�/240.

Young, H.D., 1962. Statistical Treatment of Experimental

Data. McGraw-Hill, New York, pp. 96�/101.

M.K. Dowd / Industrial Crops and Products 18 (2003) 67�/7676