Efectul Formei Si Duritatii Graului Asupra Analizei Granulometrice de La Prima Srotare a Graului

8/2/2019 Efectul Formei Si Duritatii Graului Asupra Analizei Granulometrice de La Prima Srotare a Graului

1/17

ON PREDICTING ROLLER MILLING

PERFORMANCE VI

Effect of Kernel Hardness and Shape on theParticle Size Distribution from First Break

Milling of Wheat

G. M. Campbell, C. Fang and I. I. Muhamad

Satake Centre for Grain Process Engineering, School of Chemical Engineering and Analytical

Science, The University of Manchester, UK.

Abstract: Models based on the breakage equation for roller milling have been developed topredict the output particle size distribution delivered by First Break roller milling of wheat fromdistributions of single kernel characteristics. These models allow prediction of the breakage ofmixtures of kernels of unknown origin or varieties and varying in size and hardness, basedsolely on Perten Single Kernel Characterisation System (SKCS) characteristics. Predictionshave been developed for both Sharp-to-Sharp and Dull-to-Dull roll dispositions, and showgood agreement with independent data. Milling under a Dull-to-Dull disposition is more sensitiveto kernel hardness and gives a more pronounced U-shaped distribution of output particle sizes(i.e., large proportions of both small and large particles, with few in the mid-size range) thanSharp-to-Sharp milling. Similarly, softer wheats break to give a more U-shaped distributionthan harder wheats. These findings also demonstrate that kernel hardness as reported by theSKCS is meaningful in relation to wheat breakage during roller milling. Previous work hasshown that single kernel moisture measurements can be included in predictive equations; furtherwork reported here demonstrates the potential to add the fourth SKCS parameter, kernel mass,

to predictions in order to allow for the effect of kernel shape on breakage.

Keywords: wheat; flour; hardness; roller milling; breakage equation; single kernelcharacterization.

INTRODUCTION

Flour millers produce mainly for bakers,whose principal requirement is for a flour ofconsistent quality; maintaining uniformity offlour quality was described by Scott (1951,p. 22) at the millers golden rule. The major

tool employed by millers to deliver consistentflour quality is gristing (Scott, 1951, pp. 2125; Morris, 1992; Jones, 2001, pp. 5859).Pyler (1973, p. 299) highlights this emphasison consistency and the contribution that grist-ing makes: This is a vital operation . . . sincecorrect blending of wheats constitutes thebasis for the uniformity of flour performancein the bakery. Lockwood (1945, p. 110)adds It is . . . essential that the characteristicsof a given flour should be consistent in spite ofany changes in the grist that may have to bemade from time to time owing to local or sea-

sonal market conditions. . .

. Catterall (1998)and Webb and Owens (2003) similarly echo

the consistency theme, the latter addingThe challenge for millers is to achieve thiswhile maintaining acceptable performancefrom the mill. Current surveys of bakers toidentify their top priorities for technologicaldevelopments show that constant rawmaterial (i.e., flour) quality is still the number

one priority (Sharp, 2004, personal communi-cation). At the other end of the supply chain,breeders aspire to develop wheat varietiesof consistently uniform quality (Wrigley,2002), thereby to add their contribution tothe bakers most pressing requirement.

The problem for millers is that, despite thebest efforts of breeders and growers, wheatentering the mill is inherently variable (Pyler,1958, pp. 3233; Whitworth, 1999). Kernelsfrom the same wheat stalk vary as a resultof their position on the spike during growth,and there will be variation even across a

single field. More significantly, a given varietywill be grown in different locations, under

7 Vol 85 (C1) 723

Correspondence to:

Dr G.M. Campbell, Satake

Centre for Grain Process

Engineering, School of

Chemical Engineering and

Analytical Science, The

University of Manchester, PO

BOX 88, Manchester, M60

1QD, UK.

E-mail: grant.campbell@

manchester.ac.uk

DOI: 10.1205/fbp06005

09603085/07/$30.00 0.00

Food and Bioproducts

Processing

Trans IChemE,

Part C, March 2007

# 2007 Institution

of Chemical Engineers


2/17

different agronomic practices and experiencing differentweather conditions, and many varieties are grown eachyear. Wheat is also traded internationally, with millers regu-larly blending home grown and imported wheats, particularlyin the UK (Storck and Teague, 1952, pp. 232, 266, 275;Jones, 2001, pp. 23, 52, 5859), less so in the US (Pomer-anz and Williams, 1990). The grist entering the flour milling

process is therefore likely to comprise a mixture of at leasthalf a dozen wheat varieties with widely varying origins andhistories, each inherently variable and the mixture evenmore so. The daily challenge for the miller is to cope at apractical level with the inherent variability of wheat to delivera consistent product. Their degree of technical success inthis endeavour is directly related to their commercial successin satisfying and retaining their baker customers.

Catterall (1998) notes that each wheat in a grist will haveits own peculiar characteristics, depending on variety orsource . . . [which will] affect the way the grain behavesthrough the mill, so that if the grains are blended beforemilling, the mill settings will have to be a compromise

between the various milling characteristics. (He goes on tonote, therefore, the advantages of milling varieties separatelyand blending the flours after milling; however this is expen-sive, so gristing wheat prior to milling is still the mainstreampractice.) Catteralls observation underlines the point thatgristing to achieve constant composition is not the wholestorythe resulting flour composition may be constant, butthe routes that flour takes through the mill affect other par-ameters of flour quality and mill performance, includingextraction rate and starch damage. The final flour is producedby blending together the flour produced at each milling andsifting stage, but those daughter flours have different historiesand therefore different properties. So in addition to the bulkcomposition of the grist, the various itineraries of the flour

stocks as they move through the mill also affect final flourquality. And these routes through the mill depend on the initialbreakage characteristics of the grist. Knowing these break-age patterns, in addition to the overall chemical compositionof the grist, is therefore a key factor contributing to the pro-duction of flour of consistent quality.

The purpose of flour milling is to separate bran and germfrom endosperm so as to extract as much flour as possible,at minimum operating cost, while maintaining high and con-sistent flour quality. The wheat kernel contains typically85% endosperm, so a perfectly efficient mill could in theoryextract 85% pure white flour. However, in practice, as themaximum theoretical extraction rate is approached, bran con-

tamination increases rapidly, so commercial mills tend tooperate to extraction rates in the range 7080%.Figure 1 shows a simplified diagram of a typical flour

milling process employing the gradual reduction system toachieve the goal of high recovery of flour with minimal brancontamination. The break system comprises, in this example,four pairs of rolls with their accompanying sifters, and servesto separate bran from endosperm, with some flour producedat each stage. The endosperm material that is still too large tobe considered flour is then sent to the reduction system,which serves to reduce the size of these endosperm par-ticles, again producing some flour at each stage. (Breakrolls are fluted with an asymmetric saw tooth profile and oper-ate with a gap between the rolls. The fluting breaks open the

wheat kernel such that the bran tends to stay relatively intactin large particles, while the endosperm shatters into small

particles, facilitating the separation of endosperm from branby sifting. Reduction rolls are smooth and operate underpressure, which causes damage to starch granules andthereby affects the water absorption properties of the flour;managing the degree of starch damage in the compositeflour is therefore one of tasks of the miller.) In this example,flour is produced at 21 different points in the mill, with

each flour stream having different characteristics and proper-ties. The relative proportions of these streams, and thereforethe properties of the final bulk flour, depend on the breakagecharacteristics of the grist and on the mill operation.

From Figure 1 it can be seen that the particle size distri-bution produced from First Break, the first roller milling opera-tion that the wheat encounters, determines the balance ofstream flows through the rest of the milling process. FirstBreak is therefore a critical control point in milling, the import-ance [of which] cannot be overemphasised (Hsieh et al.,1980). Yuan et al. (2003) add: There is a need for moreunderstanding of the milling system from a unit operationsapproach by determining the effect of the input stock charac-

teristics on the resulting particle size distribution of the stockleaving a break subsystem. If the miller could achieve a con-stant particle size distribution from First Break, in the face of avariable feed, then to a first approximation, the rest of the millwould run under constant conditions of flowrate to eachmilling/sifting stage. The performance of each of thesestages would then be (relatively) constant, giving more con-stant properties and proportions of the various flour fractions,and more consistent quality of the combined flour.

Delivering a constant output particle size distribution fromFirst Break requires (1) the facility to characterize the distri-bution of kernel properties in the grist; and (2) the ability topredict the breakage of the mixture during First Break millingas a function of these distributions of properties. The Single

Kernel Characterisation System (SKCS), developed by theUSDA Agricultural Research Service and commercializedby Perten Instruments, Sweden, addresses the first ofthese prerequisites. The SKCS crushes usually 300 indivi-dual kernels within 5 min and reports the distributions oftheir weight, diameter, hardness and moisture content(Martin et al., 1993; Martin and Steel, 1996; Gaines et al.,1996; Osborne et al., 1997; Ohm et al., 1998; Sissonset al., 2000; Osborne and Anderssen, 2003). The secondchallenge remains, to relate these distributions to breakageduring First Break roller milling.

Of the various quality characteristics of wheat, includingthose measured by the SKCS, hardness is the most import-

ant. Pomeranz and Williams (1990), in a comprehensivesurvey of wheat hardness research up until that time, noteKernel texture is the most important single characteristicthat affects the functionality of a common wheat . . . a para-meter of great significance in both the wheat and flour indus-try and in domestic and world trade, concluding wheat kerneltexture affects every aspect of wheat functionality exceptgluten strength and its associated factors. As hardnessincreases, so do the energy consumed in milling, the flourgranularity, damaged starch, water absorption, and both thetotal and incremental gassing power. Ohm et al. (1998)suggested that uniformity of single kernel hardness is desir-able for good milling performance. Wheat hardness is a gen-etic trait with its biochemical basis in the adhesion between

starch granules and protein in the endosperm mediated byfriabilin or puroindoline proteins (Simmonds et al., 1973;

Trans IChemE, Part C, Food and Bioproducts Processing, 2007, 85(C1): 723

8 CAMPBELL et al.


3/17

Greenwell and Schofield, 1986; Pomeranz and Williams,1990). Understanding the origins of wheat kernel hardnessand its relation to milling and baking performance has beena major driver of wheat research and a theme that pervadesthe vast majority of the wheat literature.

Previous work from our group introduced the breakageequation that describes roller milling of wheat grains(Campbell and Webb, 2001) and derived the form of thebreakage function that describes the particle size distributionresulting from breakage of an individual wheat kernel as a

function of kernel size and roll gap (Campbell et al .,2001a), from which the breakage of a mixture containing adistribution of kernel sizes could be predicted. Bunn et al.(2001) demonstrated that this form of the breakage functionwas adequate for a wide range of wheat varieties. Fangand Campbell (2003a) investigated the effect of roll disposi-tion on the breakage function, and Fang and Campbell(2003b) added a term to account for kernel moisture contentas well as size, thereby accounting for two of the four SKCSparameters. The current paper completes this work by inves-tigating the effects on breakage of the two remaining SKCSparameters, hardness and mass (which relates to kernelshape). The potential to construct a universal breakageequation that allows prediction of the breakage of an

unknown mixture of wheat kernels directly from SKCS datais thereby demonstrated.

THEORY

The breakage equation for roller milling of wheat in termsof kernel diameter is given in its cumulative form by:

P2(x)

D1D0

B(x, D)r1(D)dD (1)

where r1(D) is the particle size distribution of the feed enter-

ing the roller mill, P2

(x) is the cumulative particle size distri-bution of the output, and B(x,D) is the cumulative breakagefunction describing the proportion of material smaller thansize x in the output, originating from an input particle initiallyof size D. (See Appendix I for a discussion of the lower limit ofintegration in this equation.) In the case of wheat kernels,breakage is determined by the ratio of kernel thickness (thethird longest dimension, which equates to the diameterreported by the Perten SKCS) and roll gap, called the millingratio, G/D. Fang and Campbell (2003a) concluded that acumulative breakage function that was quadratic in both xand G/D was adequate to describe breakage under aSharp-to-Sharp (S-S) milling disposition, while under Sharp-to-Dull (S-D), Dull-to-Sharp (D-S) and Dull-to-Dull (D-D) dis-

positions, a cumulative breakage function cubic in x andquadratic in G/D was necessary. For S-S, the cumulative

Figure 1. Typical flour milling flowsheet with four break rolls. Reprinted with permission from Campbell et al. (2001b).


EFFECT OF KERNEL HARDNESS AND SHAPE ON SIZE DISTRIBUTION 9


4/17

breakage function took the form:

B(x, D) a0 b0x c0x2 (a1 b1x c1x

2)G

D

(a2 b2x c2x2)

G

D

2(2)

leading to

P2(x)

D1D0

a0 b0x c0x

2 (a1 b1x c1x

2)

G

D

(a2 b2x c2x

2)G

D

2r1(D)dD

a0 a1G1

D

a2G

2 1

D2

!

b0 b1G1

D

b2G

2 1

D2

!x

c0 c1G1

D

c2G21

D2 !

x2 (3)

where

1

Dn

D1D0

1

Dnr1(D)dD %

XNi1

pi1

Dni(4)

Nis the number of discrete size fractions into which kernels areseparated, and pi is the proportion of kernels in size fraction i.Thus, knowing (1=D) and (1=D2) and the nine coefficients ofequation (2) for the particular wheat sample, the particle sizedistribution of the output from First Break can be predicted forany roll gap. For S-D, D-Sor D-D milling, thecoefficients included0d2, increasing the total number of coefficients to 12.

Extension of the breakage equation to include SKCS hard-ness, H, gives:

P2(x)

H1H0

D1D0

B(x,D,H)r1(D)r1(H)dDdH (5)

where B(x,D,H) is theextendedbreakage function describing theproportion of material smaller than sizexproduced by breakageof an inlet particle originally of size D and hardness H. Equation(5) assumes that there is no interaction between kernel size andhardness with respect to their effects on breakage. The objec-tives of the current work were to establish the appropriate form

of the extended breakage function to include H and to confirmits predictive ability, and then to investigate the contribution ofthe fourth SKCS parameter, kernel mass.

MATERIALS AND METHODS

Milling of Wheat Samples Varying in Hardness

Nineteen wheat samples sourced from around the UK (withone, CWRS, originating from Canada) covering a range ofSKCS hardness values were milled at different roll gaps in theSatake STR-100 test roller mill, in order to generate data fromwhich to determine the form of the extended breakage function,B(x, D, H). Four additional wheat samples, along with two50 : 50

mixtures of these, were also milled to provide independent dataagainst which to validate predictions based on the extended

breakage function. The effect of moisture content has beenestablished previously (Fang and Campbell, 2003b) and wasnot included in the current study; all wheat samples were con-ditioned to 16% moisture (wet basis) and tempered overnightprior to milling. Each wheat sample was tested in the PertenSKCS (Perten Instruments AB, Sweden) before conditioning,in order to determine the initial moisture content and hence the

amount of moisture to be added, and again after conditioning.The conditioned samples were milled in 100 g batches

using the Satake STR-100 test roller mill (Satake Corporation,Hiroshima, Japan) under both S-S and D-D dispositions, usingFirst Break fluted rolls with 4.13 flutes cm21 (10.5 flutes perinch), and using six roll gaps: 0.3, 0.4, 0.5, 0.6, 0.7 and0.8 mm. Roll gaps were set using feeler gauges. Rolls wereoperated with fast and slow roll speeds of 600 and 222 rpm,respectively, giving a differential of 2.7. The mill was operatedat a feed rate of approximately 375500 kg h 21, correspond-ing to 37505000 kg m21 h21 on a full-length commercial mill.

The entire milled stocks from each trial were collected forsieve analysis using a Simon plansifter operating at 190 rpm

with a throw of 7.5 cm, and using 200 mm diameter wire meshsieves of size 2000, 1700, 1400, 1180, 850, 500 and 212 mm,along with a bottom collecting pan. A curvi-triangular sieve clea-ner was placed in each sieve to facilitate particle separation.Samples were sieved for 5 min, and the amount remaining oneach sieve measured to 0.01 g using an Ohaus PrecisionStandard balance (Ohaus Corporation, USA).

Fitting and Validation of the ExtendedBreakage Equation

For each of the test samples, the coefficients of equation(2) or its cubic equivalent were determined by least squaresregression using Microsoft Excel (Microsoft Corporation,

USA). For this purpose, (1=D) and (1=D2) were calculatedusing equation (4) and dividing the SKCS diameter datainto 20 divisions of 0.25 mm intervals covering the range16 mm. The variation of the individual coefficients with aver-age SKCS hardness was then examined. From this, anextended breakage function using a single set of coefficientsfor all the wheat samples was deduced, and its adequacyevaluated by comparing the resulting predictions againstthe independent data from the validation samples.

Residual Analysis of Remaining Variationin Relation to Kernel Shape

The departure of the actual particle size distribution for

each of the test samples from that yielded by the generalextended breakage function determined above was calcu-lated using a residual analysis. The results were consideredin view of kernel shape, to determine whether there wasevidence that kernel shape was the origin of further variationsin particle size distribution on breakage, and whether therewas merit in attempting to add the fourth SKCS parameter,mass, to the extended breakage equation (as the ratio ofkernel mass to diameter-cubed is indicative of kernel shape).

RESULTS AND DISCUSSION

SKCS Characteristics of Wheat Samples

Table 1 lists, in order of increasing average hardness, theSKCS data averaged from 300 kernels for each of the 19 test


10 CAMPBELL et al.


5/17

samples after conditioning to 16% moisture. The results for thefour validation samples, along with the two 50 : 50 blends, arealso presented. (Note that although three of the validation

samples are of the same variety as some of the test samples,they were from different origins, as reflected in their differentSKCS data.) The average SKCS hardness values range from11.2 to 80.4, so are representative of the range of wheatslikely to be encountered routinely in commercial milling. Moist-ure contents were in all cases within 0.5% of the target of16%. The smallest kernels were from the Canadian CWRSsample, with an average diameter of 2.39 mm and averageweight of 31.13 mg, while Cadenza had the largest kernels,averaging 3.35 mm in diameter and 56.32 mg in weight.Figure 2 illustrates the SKCS data for the 50: 50 mix of Consort(a soft wheat) and Spark (a hard wheat), showing the widerange of kernel sizes and the bimodal distribution of SKCS

hardness.

Fitting the Extended Breakage Function

The coefficients of the breakage function (a0, b0, . . . , c2, d2)were fitted to the data for each wheat variety milled under eachdisposition. Figure 3 shows the variation of each coefficientwith SKCS hardness following milling under the Dull-to-Dulldisposition. In each case there is a clear linear trend, implyingthat each coefficient can be described by, for example:

a0 a01 a02H (6)

and so on for the other coefficients. Thus SKCS hardness, H,

has a linear effect on breakage and can be incorporated intothe breakage function at the cost of a doubling of the number

of coefficients. The same pattern emerged from the results forS-S milling. However, inspection of this larger data set thanwas available in our previous work (Fang and Campbell,

2003a) led to the conclusion that, as for D-D milling, a cumula-tive breakage function that was cubicinxwas more appropriateforthe S-S datathana functionquadraticinxas usedpreviously.While S-S milling does give noticeably less curvature to thebreakage function than D-D milling, there seems little merit inremoving the extra degrees of freedom for describing thecurvature of the S-S data for the sake of reducing the numberof coefficients, when the larger number is required for D-Dmilling anyway. Using the same form of the breakage functionalso facilitates comparing S-S and D-D milling.

Having established this linear trend, and having come to thedecision to use cubic functions in x for both roll dispositions,extended breakage functions of the form:

B(x, D, H) (a01 b01x c01x2 d01x

3)

(a02 b02x c02x2 d02x

3)H

(a11 b11x c11x2 d11x

3)G

D

(a12 b12x c12x2 d12x

3)G

D

H

(a21 b21x c21x2 d21x

3)G

D

2

(a22 b22x c22x2 d22x

3)G

D

2H (7)

were fitted to all the data by linear regression, in each case to atotal of sixroll gaps 19 wheat varieties seven size fractions

Table 1. SKCS characteristics of wheat samples used to develop the breakage function, and of validation samples.

SKCS parameter

Mass (mg) Diameter (mm)Moisture content

(%) Hardness

M/D 3

(mg mm23)Wheat Mean SD Mean SD Mean SD Mean SD

Test samplesConsort 44.51 13.08 2.85 0.61 16.02 0.33 11.16 13.85 1.923Claire 40.81 10.81 2.66 0.59 16.25 0.33 24.59 13.25 2.168Riband 50.95 14.86 3.08 0.68 15.57 0.33 27.62 19.16 1.744Drake 46.36 14.92 2.83 0.67 15.72 0.27 29.18 15.30 2.045Crofter 51.06 16.09 2.98 0.70 15.89 0.30 41.97 15.51 1.929Soissons 44.12 12.82 2.76 0.58 15.95 0.34 52.18 15.74 2.098Raleigh 48.57 14.57 2.99 0.69 15.65 0.26 58.48 15.45 1.817Charger 45.90 11.90 2.89 0.54 15.66 0.24 59.33 16.21 1.902

Abbot 45.32 13.52 2.95 0.72 15.65 0.26 61.09 14.38 1.765Buster 52.01 11.56 3.33 0.55 15.65 0.31 62.74 17.81 1.408

Avalon 53.51 14.71 3.05 0.69 15.86 0.34 62.76 14.16 1.886Malacca 44.02 12.03 2.93 0.65 16.17 0.33 63.21 16.05 1.750Hereward 43.54 10.02 2.92 0.50 15.56 0.33 65.29 16.36 1.749Rialto 50.64 12.80 3.03 0.64 15.65 0.25 65.32 12.81 1.820Brigadier 51.03 13.89 3.02 0.65 15.51 0.29 67.14 13.87 1.853CWRS 31.13 9.10 2.39 0.56 15.91 0.40 71.48 13.73 2.280Mercia 44.28 10.87 2.90 0.66 15.54 0.23 73.63 16.49 1.816

Cadenza 56.32 12.92 3.35 0.60 15.80 0.32 76.86 14.45 1.498Spanish 45.09 14.85 2.84 0.67 16.34 0.58 80.40 21.60 1.968

Validation samplesConsort 48.87 14.54 3.06 0.65 15.79 0.27 23.85 15.83 1.706Spark 42.74 11.01 2.94 0.53 16.32 0.26 69.19 17.36 1.682Consort/Spark 44.92 12.75 2.98 0.55 15.99 0.36 47.99 26.80 1.697Malacca 46.00 12.09 3.05 0.66 15.97 0.38 61.49 13.76 1.621Soissons 44.72 13.02 2.83 0.61 15.67 0.32 54.16 14.12 1.973Malacca/Soissons 44.27 13.03 2.90 0.67 15.66 0.35 58.69 15.28 1.815




6/17

from the sieve analysis 798 data points. Table 2 lists the 24coefficients necessary to describe breakage of wheat of any

hardness under S-S milling, and the different 24 coefficientsneeded to describe D-D milling (presented in the same orderthat they appear in equation (7), to facilitate interpretation).(Note that these coefficients are dimensional, requiring xto bemeasured in micrometres, and giving results as percentages.)From these coefficients, in principle the particle size distributionresulting from First Break milling of any wheat, or any mixture ofwheats, at any roll gap in the range 0.30.8 mm can be pre-dicted directly from the distributions of kernel size and diametermeasured by the SKCS.

Table 2 also shows the ratio of the coefficients, indicatingthe effect on the breakage function of changing from S-S toD-D milling. Many of the ratios are close to unity, indicatinglittle effect of roll disposition on the particle size/hardness/

roll gap interaction indicated by those coefficients. One ofthe most striking changes is in the d02 coefficient, which

changes sign from negative to positive and increases inmagnitude by a factor of 8.8. The large positive coefficient

at this point indicates that the cumulative distribution tendsto change to one that begins to curve more, rather thanless, steeply at the extreme particle sizes, indicating a ten-dency towards greater quantities of both large and small par-ticles under D-D milling compared with S-S. This is borne outby the results discussed below. The location of this coefficientalso implies that the extent of this effect is dependent onhardness; the difference between S-S and D-D milling ismore dramatic for hard wheats than for soft. By contrast,most of the other an2 coefficient ratios are positive andclose to unity, indicating little change in the (G/D)2H or(G/D)22H interactions between S-S and D-D milling.Beyond this, the coefficients interact and compensate soextensively that it is difficult to disentangle specific effects.

It is interesting to note, however, that the a11 coefficientratios are all changed substantially, with a11 increasing by a

Figure 2. SKCS data for a 50 : 50 mixture of Consort (a soft wheat) and Spark (a hard wheat).


12 CAMPBELL et al.


7/17

factor of 6.71 and b11, c11 and d11 decreasing by factors ofaround 40, 12 and 9, respectively, the latter two also chan-ging sign. This implies a significant difference in the effectof G/D between S-S and D-D milling.

It ought to be possible, with further investigation, to reducethe number of coefficients required to describe the effects ofkernel size, hardness, roll gap and roll disposition on break-age. Fewer coefficients would reduce the quantity of

Figure 3. Variation with SKCS hardness of the coefficients in the D-D breakage function for individual wheat samples.

Table 2. Coefficients in the breakage functions for S-S and D-D milling, and ratios between corresponding coefficients.

n an1 bn1 cn1 dn1 an2 bn2 cn2 dn2

S-S disposition0 23.176 1.613 1021 21.049 1024 2.431 1028 25.310 1021 1.271 1023 24.768 1027 23.735 10211

1 23.342 101 24.750 1021 5.559 1024 21.518 1027 7.463 23.152 1022 2.761 1025 26.474 1029

2 1.479 102 3.274 1021 29.198 1024 3.204 1027 21.899 101 8.692 1022 28.619 1025 2.282 1028

D-D disposition

0 1.612 101 1.722 1021 21.265 1024 3.217 1028 27.471 1021 2.210 1023 21.623 1026 3.287 10210

1 22.241 102 21.202 1022 24.756 1025 1.612 1028 7.119 23.247 1022 3.230 1025 28.413 1029

2 5.216 102 21.028 1.003 1023 22.719 1027 21.467 101 7.581 1022 28.388 1025 2.365 1028

Ratio D-D : S-S

0 25.076 1.068 1.206 1.323 1.407 1.739 3.404 28.8011 6.706 0.025 20.086 20.106 0.954 1.030 1.170 1.3002 3.527

23.140

21.090

20.849 0.773 0.872 0.973 1.036




8/17

experimental data needed to fit the equation and wouldsimplify the industrial application of the breakage equationfor on-line control of First Break milling.

Yuan et al. (2003) described the particle size distributionfrom the five break systems of a pilot flour mill using the logis-tic equation (adapted to be consistent with the currentnomenclature):

P2(x) eabx

c

1 eabxc 100 (8)

For First Break, they related the three parameters of thisequation to wheat characteristics including test weight,thousand kernel weight, ash content, protein content andSKCS hardness, mass, diameter and moisture content, andto subsets of these characteristics, based on milling of ninecommercial blends of hard red winter wheat from the19971998 Kansas harvests. These yielded empiricalequations with 12 18 fitted coefficients. For subsequentbreaks, the mean diameter and standard deviation of the

input stocks to each break were also included in theequations, requiring 11 15 fitted coefficients. Predictionsagainst independent data from a tenth batch were reasonablefor First Break but poor for the subsequent breaks.

A limitation of equation (8) is that it is only able to describecumulative particle size distributions that conform to a sigmoi-dal profile. While this is generally appropriate for hard wheatsmilled under a S-S disposition (see later), it does not have theflexibility of equation (7) to describe other typical psds result-ing from milling. In particular, soft wheats and D-D millingtend to give cumulative psds after First Break milling thatare inversely sigmoidal, indicative of greater proportions ofparticles at the extremes and fewer particles in the mid-sizeranges. Equation (8) also makes the relationship between a

particular wheat characteristic and its effect on breakagevery indirect and difficult to interpret. For example, for FirstBreak milling, SKCS hardness features in the equations forboth the a and b parameters in equation (8). This makes itdifficult to conclude, for example, that kernel hardness hasa linear effect on breakage, as we are able to do in the currentwork.

Al-Mogahwi and Baker (2005) investigated breakage inboth break and reduction roll systems in a commercial mill,and suggested some alternative approaches for characteri-zing the particle size distributions of flour stocks from theseoperations. Their approaches might allow simpler forms ofthe breakage function to be developed, as well as facilitatingextension of the breakage equation approach developedhere for First Break to the rest of the milling process. Theirequation

P2(x) 1 1x=G

k

exp((x=G)=(k))

100 (9)

allows the effects of particle size x and roll gap G on theparticle size distribution to be described by a single para-meter k, which could dramatically reduce the number ofcoefficients required to describe effects of hardness andother parameters on breakage. However, equation (9) sufferseven more than equation (8) in terms of limited flexibility todescribe the range of psds encountered in wheat flour

milling. The effects of roll disposition and of kernel size,moisture and hardness on k have not been investigated.

Effects of Hardness on Breakage During FirstBreak Roller Milling

Figure 4 confirms the linear effect of hardness by plottingthe percentage smaller than x versus average kernel hard-ness, for each aperture size used in the sieve analysis, andfor each roll gap. The fitted extended breakage functions

are also shown as solid lines (although note that the fitwould be slightly different for each wheat sample at a givenroll gap, as the average kernel size differs between samples;the fits shown are therefore based on a typical averagekernel size). Close inspection of this figure clarifies the effectsof hardness and roll gap on wheat breakage and the differ-ences between S-S and D-D milling.

Increasing roll gap results in fewer small particles and morelarge particles, as expected, with D-D more sensitive to rollgap changes than S-S (which also implies greater sensitivityto kernel size, as size and roll gap have equivalent effects onbreakage). In all cases there is a divergent pattern, such thatthe percentage smaller than 2000 mm increases with hard-ness (implying the percentage of large particles decreases),

while the percentage smaller than 212 mm decreases withincreasing hardness. This demonstrates that soft wheatstend to break to produce relatively larger proportions ofboth larger and smaller particles, with fewer in the mid-sizerange, compared with hard wheats which produce fewerparticles at the extremes and more in the middle. Figure 5illustrates this with the particle size distributions resultingfrom different roll gaps under both roll dispositions forClaire, a soft wheat, and Mercia, a hard wheat. Under S-S,Claire gives a relatively straight line distribution across therange, with the slope (i.e., the relative proportions of smalland large particles) depending on the roll gap, while Merciagives a pronounced peak in the distribution, with increasing

roll gap increasing the proportion of larger particles andmoving the peak to the right. Under D-D the peak has dis-appeared for Mercia, while Claire has moved to a pro-nounced U shape with large proportions of both large andsmall particles, and few in the mid-size range. In general,D-D milling gives a more U-shaped distribution for a givenwheat hardness, in agreement with our previous findings(Fang and Campbell, 2002a, b, 2003a), while softness alsocontributes to a more U-shaped distribution. This indicatesthat soft wheats tend to shatter easily into numerous smallendosperm particles, while leaving the bran material rela-tively intact as large particles. Hard wheats, by contrast,transmit the stresses throughout the kernel, such that thebran breaks along with the endosperm, the latter resistingshattering into numerous small particles (Pomeranz andWilliams, 1990). Dull-to-Dull milling gives more of a crushingaction, which encourages shattering of the brittle endospermbut leaves the bran layers relatively intact, while the shearingaction of Sharp-to-Sharp milling cuts through both the branand the endosperm material, slicing the kernel into smallerparticles but not shattering it to the same extent as Dull-to-Dull. Thus wheat hardness and roll disposition have similareffects, such that a soft wheat such as Claire under D-Dgives a pronounced U-shape, while at the other extreme, ahard wheat under S-S gives quite a peak or an inverted U.

It is also clear that the divergent patterns of Figure 4 aremore pronounced under a Dull-to-Dull disposition than

under Sharp-to-Sharp. This indicates that D-D milling ismore sensitive to wheat hardness than S-S. As millers tend


14 CAMPBELL et al.


9/17

Figure 4. Percentage smaller than xversus average SKCS hardness for different wheats milled under S-S (left) and D-D (right) roll dispositions,at roll gaps of 0.3 mm, 0.4 mm, 0.5 mm, 0.6 mm, 0.7 mm and 0.8 mm.




10/17

Figure 4. Continued


16 CAMPBELL et al.


11/17

to operate under D-D (because the U-shaped distribution isreadily separated into larger branny particle and smaller

endosperm particles), the greater sensitivity of D-D millingto kernel hardness and size is commercially significant, asvariations in the feedstock will have greater influence ondownstream operations.

The consistent linear trends in Figure 4 imply that the effectof hardness on breakage is qualitatively similar for wheats ofvarying hardness; there are no sudden discontinuities thatwould indicate that hard and soft wheats are different intheir essential nature or exhibit fundamentally different mech-anisms of breakage. This agrees with Osborne and co-workers conclusions based on SKCS crush responseprofiles, that wheat breakage patterns have a common gen-eric structure, in contrast with those of barley (Osborne and

Anderssen, 2003; Osborne et al., 2004). The linear effect

also implies that the average SKCS hardness, rather thanthe full hardness distribution, is sufficient to predict breakage

at First Break, even for mixtures of wheats of very differenthardnesses. This is in contrast to the effect of kernel dia-

meter, for which the inverse quadratic relationship with break-age requires measurement of the entire diameter distribution.The results also indicate that SKCS hardness is meaning-

ful with respect to breakage during roller milling. This is a sur-prising finding, as the breakage mechanism in the SKCS,involving a single rotor with a relatively fine sawtooth profilecrushing kernels against a smooth stationary crescent witha large gap between (Martin and Steele, 1996), is very differ-ent from the breakage action during First Break roller milling.Muhamad and Campbell (2004) showed that the SKCS givesvery similar particle size distributions for wheats of differenthardness, implying that the SKCS hardness index primarilyreflects the energy to grind to a consistent degree of break-age, and is therefore indicates a fundamental property of

the wheat kernel, rather than an artefact of the particularbreakage system. This is probably the reason that SKCS

Figure 5. Particle size distributions from Claire (a soft wheat, SKCS hardness 24.6) and Mercia (a hard wheat, SKCS hardness 73.6) atdifferent roll gaps under S-S (left) and D-D (right) roll dispositions.




12/17

hardness relates so well to breakage during First Breakroller milling.

Validation of the Extended Breakage Function

Figures 6 and 7 show the predictions of the particle sizedistributions resulting from breakage of the six validation

samples (four pure varieties and two mixtures) at differentroll gaps and under S-S and D-D dispositions. (For clarity,the data for only three roll gaps is presented.) Clearly, theagreement is excellent in most cases, particularly for Consortand Spark and their mixture, with poorest agreement forSoissons, and generally poorer agreement at the larger rollgap. Overall the good agreement indicates that includingSKCS hardness in the breakage function has taken it wellon the way to being more universal; from equations (5) and(7), and using the coefficients in Table 2, the entire particlesize distribution resulting from breakage of any mixture ofwheat kernels of any distribution of size and hardness, atany roll gap and under either S-S or D-D disposition, can

be predicted with good accuracy. This is a powerful develop-ment that allows information from the SKCS to be useddirectly to predict breakage during roller milling, therebyallowing the miller to know how todays particular grist islikely to behave on the mill, and to adjust the mill settingsaccordingly. One can readily envisage a control system(based, for example, on the approaches of Wang et al.,2005) that would take off-line or, with some adjustment, on-line SKCS data and feed it forward to control First Break auto-matically to deliver a consistent output particle size distri-bution, or to alter the downstream processing in responseto a predicted change to the output from First Break.

Effect of Kernel Shape on BreakageDespite the clear success of adding hardness to the break-

age function and the opportunities this presents, it is evidentfrom Figures 6 and 7 as well as from Figure 4 that there aresome varieties that exhibit anomalous breakage patterns andfor which the prediction is not so accurate. This may be dueto additional factors such as kernel shape. Kernel shape isrelated to mass; a long thin kernel will have a greater massthan a short kernel of the same thickness (assuming shapedifferences to be more significant than density differences).However, mass increases with D 3 for the same shape anddensity. Thus the ratio of SKCS mass to D 3 is an indicationof kernel shape; a larger ratio corresponds to relatively

more elongated kernels. The extended breakage function ofequation (7) includes kernel hardness and diameter asmeasured by the SKCS, while kernel moisture can beadded as described previously (Fang and Campbell,2003b). Adding mass would introduce the fourth SKCS para-meter into the breakage function and would allow kernelshape effects to be accounted for.

In order to determine whether mass and shape effects didindeed account for some of the remaining variation inbreakage patterns, a residual analysis on the data from the19 test varieties was performed. Each variety was milled atsix different roll gaps, giving six independent results for theagreement between the predicted and experimental particlesize distributions. If all (or a statistically significant majority)

of the differences were of the same sign (positive or nega-tive), this would indicate a systematic error in the prediction

for this variety. Whether the systematic error appeared tobe related to kernel mass could then be investigated. Fulldetails of this study are given by Muhamad (2004). (Notethat Muhamad performed the study using a breakage func-tion for S-S milling that was quadratic in x, not cubic asused here. The following results are as reported byMuhamad, but would not be substantially different or lead

to different conclusions had a cubic in x function been usedfor the S-S analysis.)

The basis of the analysis was that, for a given roll gap, thepredictive model would tend to either underpredict or overpre-dict the cumulative particle size distribution over the entirerange, such that the residuals at each point would tend tobe of the same sign. Therefore the residuals at each pointwere summed in order to establish an overall residual foreach roll gap:

Total residual X8i1

(measured P2(xi) fitted P2(xi))

(10)

Figure 8 shows the total residual for the six roll gaps foreach variety, plotted against average kernel mass for thatvariety, for both roll dispositions. Clearly (and confirmed by

Analysis of Variance) in some cases the deviation of theaverage residual for a variety was significantly differentfrom zero, compared with the within-variety variation, inother words, the deviation was systematic for particular var-ieties rather than a result of random experimental error.Figure 8 shows that for kernels that were neither exception-ally heavy nor exceptionally light, mass had no discernibleeffect on deviations of predictions from experimental break-age. However, where the systematic difference was substan-

tial, the reason clearly relates to the kernel mass. The twoextreme points under both roll dispositions correspond toCWRS with the lowest kernel mass and large positiveresiduals, and Cadenza with its large kernels and large nega-tive residuals. Kernel size has already been taken into con-sideration in the breakage function, so the very large andvery small masses of Cadenza and CWRS, respectively, indi-cate that shape effects are likely to be contributors to the rela-tively poorer predictions obtained for these two varieties.[Differences in kernel density rather than shape could bethe reason for the different breakage patterns. Density isslightly correlated with kernel hardness (Pomeranz and Mat-tern, 1988; Fang and Campbell, 2000; Dobraszczyk et al.,

2002), and these two wheats have similar hardness valuesbut very different shapes, so it is likely that the observed devi-ations are related to kernel shape rather than density.] Longthin kernels would be heavier relative to their thicknessthan short fat kernels (assuming negligible density effects),such that the ratio of SKCS mass : diameter3 (M/D 3) oughtto reflect kernel shape differences. Table 1 lists the SKCSM/D 3 ratios for the different varieties. CWRS had an M/D 3

value of 2.280 mg mm23, the highest value found in thesample set, suggesting a relatively long kernel for its size,while Cadenzas value of 1.498 mg mm23 was the thirdlowest, suggesting that shape was indeed a major contribu-tory factor in the dissimilar breakage of these samples com-pared with others. Figure 9 illustrates the differences

between the kernels of Cadenza and CWRS used in thestudy. While acknowledging that the data set is dominated


18 CAMPBELL et al.


13/17

Figure 6. Comparison of predicted (lines) and experimental (symbols) cumulative particle size distributions for Consort, Spark and a 50 : 50mixture of Consort and Spark milled under S-S (left) and D-D (right) roll dispositions.




14/17

Figure 7. Comparison of predicted (lines) and experimental (symbols) cumulative particle size distributions for Mallacca, Soissons and a 50 : 50mixture of Mallacca and Soissons milled under S-S (left) and D-D (right) roll dispositions.


20 CAMPBELL et al.


15/17

by these two extreme points, it is noteworthy that of the vali-dation samples, predictions were poorest for Soissons which,like CWRS, tended to be underpredicted and which alsoexhibited a relatively high M/D 3 ratio of 1.973 mg mm23.

The shape of wheat kernels is complex, and the sample setused here small and of limited diversity of origin, so it is per-haps not surprising that clear shape correlations were elu-sive. The additional complexity of including kernel mass inthe breakage function is probably not merited for mostwheats; however, the results here show that it could bedone and may be worthwhile for kernels of unusual size orshape.

Single Kernel Characteristics in Relation toMilling and Baking Performance

The results presented in the current work demonstrate apractical basis for exploiting SKCS measurements toenhance and control mill performance. First Break is a criticalcontrol point in flour milling, with the particle size distribution

created at that point determining the flows through the restof the mill and the combination of processing historieswithin the final flour. The breakage equation approach devel-oped here could be extended to second and subsequentbreaks and to reduction roller milling. The results also givea sound basis for beginning to interpret other aspects ofmilling and baking performance in terms of mechanistic con-nections (as opposed to the statistical correlations that ratherdominate the literature). For example, the particle size distri-bution at First Break, and indeed at subsequent break andreduction rolls, can be related to energy consumption (Pujolet al., 2000), which is a major cost in flour milling. The relationbetween wheat hardness and flour yield, so often noted (e.g.,Pomeranz and Williams, 1990; Ohm et al., 1998), dependson the initial and subsequent breakage characteristics ofthe grist, for which the breakage equation allows quantitativedescription. In addition to particle size, other aspects of flourfunctionality such as particle composition or starch damagecould be incorporated into breakage equation models(Fistes and Tanovic, 2006). This would facilitate in makingthe entire series of linkages between wheat hardness, millingperformance and baking consistency.

CONCLUSIONS

The breakage function for roller milling has been extendedto include kernel hardness as well as size, such that theentire particle size distribution resulting from First Breakmilling of a mixture of wheat varieties, of unknown originand differing in size and hardness, milled at any roll gapand under either D-D or S-S milling, can be predicted to agood first approximation directly from SKCS data. Harderwheats break to give a more even distribution of particlesthan softer wheats, which tend to break to give largeproportions of large and small particles, with fewer in themid-size range. Dull-to-Dull milling similarly gives a more U-shaped distribution than Sharp-to-Sharp milling, and ismore sensitive to wheat hardness and to roll gap changes.

Wheat varieties of differing hardness break in qualitativelythe same way, such that breakage patterns vary smoothly

Figure 8. Total residual for each of six roll gaps versus kernel mass for 19 wheat varieties milled under S-S (left) and D-D (right) dispositions.

Figure 9. Representative kernels of Cadenza and CWRS, showing

ventral (left), lateral (middle) and dorsal (right) views, illustratingsize and shape differences (bar 5 mm).




16/17

with hardness. Wheat hardness has a linear effect onbreakage, indicating that the SKCS hardness index is inher-ently meaningful in relation to commercial milling of wheat onfluted roller mills. Previous work has established that effectsof kernel moisture on breakage can also be predicted. Thecurrent work has shown that the fourth SKCS parameter,kernel mass, can also be included in the breakage function,

but that the additional complexity yields little additionalbenefit for most wheat varieties. The work has shown thatdistributions of single kernel data, as measured by theSKCS, can be used directly to predict breakage during FirstBreak roller milling, either off-line or as part of an automaticcontrol system. This would aid millers in delivering consistentquality flour to bakers in the face of a constantly varying feed-stock. The quantitative nature of the breakage equationapproach could also facilitate understanding the mechanisticrelationships between wheat kernel characteristics and thevarious facets of milling and baking performance.

NOMENCLATUREa2c coefficients in equation (8)an2dn coefficients in the breakage functionann2dnn coefficients in the extended breakage functionB(x, D) cumulative breakage function describing the

proportion of material smaller than size x in theoutput, originating from an input particle initially ofsize D

B(x, D, H) extended cumulative breakage function describingthe proportion of material smaller than size x in theoutput, originating from an input particle initially ofsize D and hardness H

D size of input particle, SKCS diameterD-D Dull- to-Dull rol l d isposit ionG roll gapH SKCS kernel hardness

k fitting parameter in equation (9)M SKCS kernel massn an integer taking values of 0, 1 or 2N number of discrete size fractions into which kernels

are separated for the purpose of calculatingaverage values

pi proportion of kernels in size fraction iP2(x) cumulative particle size distribution of the outputS-S Sharp-to-Sharp roll disposit ionx size of output particle

Greek symbolsann coefficients in the extended breakage functionr1(D) particle size distribution of the feedr1(H) hardness distribution of kernels in the feedr2(x) particle size distribution of the output

REFERENCES

Al-Mogahwi, H.W.H. and Baker, C.G.J., 2005, Performance evalu-ation of mills and separators in a commercial flour mill, TransIChemE, Part C, Food Bioprod Proc, 83: 2535.

Austin, L.G., 1972, A review: Introduction to the mathematicaldescription of grinding as a rate process, Powder Technol, 5:117.

Bunn, P.J., Campbell, G.M., Fang, C. and Hook, S.C.W., 2001, Onpredicting roller milling performance. Part III. The particle sizedistribution from roller milling of various wheats using fluted rolls,Proc 6th World Chemical Engineering Congress, University of Mel-bourne, Melbourne, Australia.

Campbell, G.M. and Webb, C., 2001, On predicting roller milling

performance. I. The Breakage Equation, Powder Technol, 115:234242.

Campbell, G.M., Bunn, P.J., Webb, C. and Hook, S.C.W., 2001a, Onpredicting roller milling performance. II. The Breakage Function,Powder Technol, 115: 243255.

Campbell, G.M., Fang, C., Bunn, P.J., Gibson, A.A., Thompson,F. and Haigh, A., 2001b, Wheat flour milling: A case study in pro-cessing of particulate foods, in Hoyle, W. (ed.). Powders andSolids Developments in Handling and Processing Technologies,95111 (Royal Society of Chemistry, Cambridge, UK).

Caterall, P., 1998, Flour milling, in Cauvain, S.P. and Young, L.S.(eds). Technology of Breadmaking, 296329 (Blackie Academicand Professional, London, UK).

Dobraszczyk, B.J., Whitworth, M.B., Vincent, J.F.V. and Khan A.A.,2002, Single kernel wheat hardness and fracture properties inrelation to density and the modelling of fracture in wheat endo-sperm, J Cereal Sci, 35: 245263.

Fang, C. and Campbell, G.M., 2000, Effect of measurement methodand moisture content on wheat kernel density measurement, TransIChemE, Part C, Food Bioprod Proc, 78: 179186.

Fang, C. and Campbell, G.M., 2002a, Stress-strain analysis andvisual observation of wheat kernel breakage during first breakroller milling, Cereal Chem, 79: 511517.

Fang, C. and Campbell, G.M., 2002b, Effect of roll fluting dispositionand roll gap on the breakage of wheat kernels during first breakroller milling, Cereal Chem, 79: 518522.

Fang, C. and Campbell, G.M., 2003a, On predicting roller milling per-

formance IV: Effect of roll disposition on the particle size distri-bution from first break milling of wheat, J Cereal Sci, 37: 2129.

Fang, C. and Campbell, G.M., 2003b, On predicting roller milling per-formance V: Effect of moisture content on the particle size distri-bution from first break milling of wheat, J Cereal Sci, 37: 3141.

Fistes, A. and Tanovic, G., 2006, Predicting the size and com-positional distributions of wheat flour stocks following first breakroller milling using the breakage matrix approach, J Food Eng,75: 527534.

Gaines, C.S., Finney, P.F., Fleege, L.M. and Andrews, L.C., 1996,Predicting a hardness measurement using the Single-KernelCharacterisation System, Cereal Chem, 73: 278283.

Greenwell, P. and Schofield, J.D., 1986, A starch-granule proteinassociated with endosperm softness in wheat, Cereal Chem, 63:379380.

Hsieh, F.H., Martin, D.G., Black, H.C. and Tipples, K.H., 1980, Somefactors affecting First Break grinding of Canadian wheat, CerealChem, 57: 217223.

Jones, G., 2001, The Millers A Story of Technological Endeavourand Industrial Success, 18702001 (Carnegie Publishing Ltd,Lancaster, UK).

Lockwood, J.F., 1945, Flour Milling(The Northern Publishing Co. Ltd,London, UK).

Martin, C.R., Rousser, R. and Brabec, D.L., 1993, Development of asingle kernel wheat characterisation system, Trans American Socof Agric Eng, 36(5): 1399 1404.

Martin, C.R. and Steele, J.L., 1996, Evaluation of rotor-crescentdesign for sensing wheat kernels hardness, Trans American SocAgric Eng, 39(6): 2223 2227.

Morris, C.F., 1992, Impact of blending hard and soft white wheats onmilling and baking quality, Cereal Foods World, 37: 643648.

Muhamad, I.I., 2004, Single kernel effects on breakage during wheatmilling, PhD thesis, UMIST, Manchester, UK.

Muhamad, I.I. and Campbell, G.M., 2004, Effects of kernel hardnessand moisture content on wheat breakage in the Single KernelCharacterisation System, Innovative Food Sci Emerging Technol,5: 119125.

Ohm, J.B., Chung, O.K. and Deyoe, C.W., 1998, Single-kernelcharacteristics of hard winter wheats in relation to milling andbaking quality, Cereal Chem, 75: 156161.

Osborne, B.G., Kotwal, Z., Blakeney, A.B., OBrien, L., Shah, S. andFearn, T., 1997, Application of the Single-Kernel CharacterisationSystem to wheat receiving testing and quality prediction, CerealChem, 74: 467470.

Osborne, B.G. and Anderssen, R.S., 2003, Review: Single kernelcharacterization principles and applications, Cereal Chem, 80:613622.

Osborne, B.G., Anderssen, R.S. and Huynh, H.-N., 2004, In-situmeasurement of the rheological properties of wheat and barleyusing the SKCS 4100, in Cauvain, S.P., Salmon, S,E. and Young,

L.S. (eds). Proceedings of the 12th ICC Cereal and Bread Con-gress, 207211 (Woodhead Publishing Ltd, Cambridge, UK).


22 CAMPBELL et al.


17/17

Pomeranz, Y. and Mattern, P.J., 1988, Genotype and genotype environmental interaction effects on hardness estimates in winterwheat, Cereal Foods World, 33: 371374.

Pomeranz, Y. and Williams, P.C., 1990, Wheat hardness: its genetic,structural and biochemical background, measurement and signifi-cance, in Pomeranz, Y. (ed.). Advances in Cereal Science andTechnology, 471544 (American Association of Cereal Chemists,St Paul, Minnesota).

Pujol, R., Le

tang, C., Lempereur, I., Chaurand, M., Mabille, F. andAbecassis, J., 2000, Description of a micromill with instrumentationfor measuring grinding characteristics of wheat grain, CerealChem, 77: 421427.

Pyler, E.J., 1958, Our Daily Bread (Siebel Publishing Company,Chicago, USA).

Pyler, E.J., 1973, Baking Science and Technology Volume I (SiebelPublishing Company, Chicago, USA).

Scott, J.H., 1951, Flour Milling Processes (Chapman and Hall Ltd,London, UK).

Simmonds, D.H., Barlow, K.K. and Wrigley, C.W., 1973, The biochemi-cal basis of grain hardness in wheat, Cereal Chem, 50: 553 563.

Sissons, M.J., Osborne, B.G., Hare, R.A., Sissons, S.A. and Jack-son, R., 2000, Application of the single-kernel characterizationsystem to durum wheat testing and quality prediction, CerealChem, 77: 410.

Storck, J. and Teague, W.D., 1952, Flour for Mans Bread(University

of Minnesota Press, Minneapolis, USA).Wang, H., Zhang, J.F. and Yue, H., 2005, Periodic learning of B-

spline models for output PDF control: application to MWD control,Proceedings of the American Control Conference 2, 955960.

Webb, C. and Owens, G.W., 2003, Milling and flour quality, inCauvain, S. (ed.). Bread Making: Improving Quality, 200219(Woodhead Publishing Ltd., Cambridge, UK).

Wrigley, C.W., 2002, Walter Bushuk: Cereal chemist and mentor, inNg, P.K.W. and Wrigley, C.W. (eds). The Bushuk Legacy, 147(American Association of Cereal Chemists, Minnesota, USA).

Whitworth, M.B., 1999, Heterogeneity in structure and grain compo-sition, Proceedings of Process Engineering of Cereal ProductsWorkshop, Montpellier, France, 8th October 1999 (published byICC, Austria).

Yuan, J., Flores, R.A., Eustace, D. and Milliken, G.A., 2003, A sys-tematic analysis of the break subsystems of a wheat flour pilotmill, Trans IChemE, Part C, Food Bioprod Proc, 81: 170179.

ACKNOWLEDGEMENTS

This work was funded in part by the UK Engineering and PhysicalSciences Research Council (EPSRC, Grant No. GR/M49939). Thesupport of the Satake Corporation of Japan is gratefully acknowl-edged, along with ADAS and Marriages Mills for assistance withsourcing wheat samples. IIM gratefully acknowledges the UniversitiTeknologi Malaysia for funding to pursue PhD studies. PrasanChoomjaihan is gratefully acknowledged for producing Figure 9.

The manuscript was received 5 January 2006 and accepted forpublication after revision 25 July 2006.

APPENDIX I: THE LOWER LIMIT OF INTEGRATIONIN THE BREAKAGE EQUATION

Campbell and co-workers original papers (Campbell andWebb, 2001; Campbell et al., 2001a) gave the lower limitsof the integration as D x, in common with previous workerssuch as Austin (1972), but Fang and Campbell (2003a, b)altered this, without explanation, to D 0. The explanationis that D and x are measuring different things and are notdirectly comparable. D is kernel thickness (measured bythe SKCS, by image analysis or using slotted sieves), while

x is the smallest square aperture through which a particlewill pass, as measured by sieve analysis. It is thereforemeaningless in this context to write D x. Previous workersstudying breakage have performed the integration from D xon the basis that they were measuring the size distributions

of their input and output material in the same way, e.g.,using the same sieve analysis procedure, and arguing thatthe breakage process implied that output particles of size xcould only have arisen from inlet particles originally largerthan x. This is a reasonable argument that simplifies themathematics. However, it renders the breakage equationless general and less flexible, by excluding the possibilitythat inlet and outlet particles might be measured differently,and that the particular dimension chosen to characterize anoutlet particle size might be larger in magnitude than thedimension used to characterize the size of the inlet particlefrom which that outlet particle originated. In the case ofwheat, the roller milling process tends to open up the initiallycompact wheat kernel to create large bran flakes along with

finer endosperm material. Thus it is quite possible for akernel, initially of 2 mm diameter for example, to yield abran flake measuring 3 mm across when measured bysieve analysis. Hence the appropriate range for the inte-gration in equation (1) is from zero to infinity. This approachalso leads naturally to extensions of the breakage equationto include factors affecting breakage in addition to inlet par-ticle size distribution, such as hardness distribution, as inthe current paper.



Efectul Formei Si Duritatii Graului Asupra Analizei Granulometrice de La Prima Srotare a Graului

Documents

Transcript of Efectul Formei Si Duritatii Graului Asupra Analizei Granulometrice de La Prima Srotare a Graului