Copyright is owned by the Author of the thesis.

Permission is given for

a copy to be downloaded by an individual for the purpose of research and
private study only. The thesis may not be reproduced elsewhere without
the permission of the Author.
 CONTINUOUS BUTTERM.AKING - A PROCESS

CAPABILITY STUDY

A thesis presented in partial fulfilment

of the requi r ements f or the degree
of Master of Technology in
Industrial tianagement at
Mass ey University

Dean Thomas John Stockwell

1972
 ii

11
If you cn.n raensure that of which you
speak , and you cnn cxpr esn it by a
number , you know soraething of your
subject, but if you carmot measure
it, your know l edge is raeagre
.:md unsatisfnctor y"

Lor d Kelvin.
 iii

A process capability study was conducted on a Contimab MC 30

continuous buttermaking machine . The compositional parameters of
butter moisture and salt content were considered.

The initial investigation showed that compositional variation

with respect to time was signifi cantly greater than variation within
the prod.ict at any one instant , A significant correlation was found
be t ween variations in moisture and salt content and it was considered
that variation in both moistur e and salt content was strongly influenced
by the variable performance of the salt slurry i n j ection system.

The preceeding results suggested examination of the salt slurry

injection pressure and linear extrusion speed of the butter ribbon.
A complex relationship was s een to exist between these factors and
the pr oduct composition; possible explanations are considered .
 iv
ACKNOWLEDGENENTS

I wish to thank the Supervisory panel Professor J.K. Scott,

Dean of the Faculty of Food Science and Biotechnology; Dr. Ro Dolby
and Mr. s. Jebson of the New Zealand Dairy Research Institutef for
their guidance during this work. I would like to include special
thanks to Mr. K. Noonan, Department of Industrial Hanagement and
Engineering 9 who has helped in innumerable ways during the project
and in particular for his help during sampling runs.

I am also grateful to Nr . a. Foot and Hr. D. Kingsbeer,

Department of Industrial Management and Engineering, for assistance
with measurement equipment and a lso during sampling runso

I am indebted to the Manawatu Cooperative Dairy Company Limited

for use of facilities during the study. In particular I would like
to thank Mr. G. :'.::. Baker, General Manager, and Mr. E. Petch, Manager
of the Butter Factory, without whose support this project would not
have been possible.

Thanks are also due to : -

Staff members of the Faculty of Food Science and Biotechnology
for assistance in many ways.
Mr. R. Russell, New Zealand Dairy Research Institute for
information with regard to continuous churn operation.
The Massey University Computer Unit for he lp in data processing.
Mr. P. Herbert of the Nassey University Printery for printing
of diagrams and figures.
Hr. R. Leitch of the Massey University Photographic Unit for
photographs contained in the thesis.

I wish to acknowledge financial support from New Zealand

Cooperative Dairy Company and Mauri Brothers and Thompson Limited.

Finally, I am most grateful to Mrs. B.R. Robertson for typing

of the script and assistance during printing of the final copy.

D. T. J . Stockwell
(November, 1972)
 TABLE OF COHT~TS

page
Acknow ledgernen ts
List of figures
List of tables
Section I
Introduction.
Literature r eview o 7
The economic importance of butt er composition . 12
The Contimab MC 30 continuous buttermaking machine. 15
(The influence of machine and raw material factors.)
Section II 26
Methods of analysis . 26
Section III Experimental \fork 35
(i) A study of we i ght changes during cooling of samples in 35
the gr avimetric analysis of moisture content of butter .
(ii) A study of variations in the product composition with 37
respect to the Contimab 11C 30 .
(iii) A study of product variation over a short time period. 44
(iv) A study of product variation over an extended time 48
period "
(v) An examina ti on of salt injec tion pump performance . 67
(vi) Frequency distribution analysis of product composition 68
data
(vii) Conclusions . 71
Section IV An i nvestigat i on of selected pro cess variables 72
(i) Raw material f a ctors . 72
- cream temperature
- cream fat content
- cream acidity
(ii) Continuous churn machine variables " 93
- linear extrusion speed
- salt pump injection pressure
(iii) Discussion c 127
(iv) Conclusions . 130
Section V Suggestions for further study 131
Section VI Conclusions 133
Appendix I Numerical results for weight changes during 1 34
cooling.
Appendix II Summary of analysis of variance for slice 135
samples.
Appendix III .Analysis of results using Cumulative Sum 138
·""
(Cusum) t e chniques.
Appendix IV Salt slurry inj ection pump performance. 142
Appendix V Normality calculation for frequency 145 ~

distribution of moisture and salt data.

Appendix VI Summary of data other than butter moisture 147
and salt values .
Appendix VII Listing of selected computer programs. 148 )>-,

Bibliography 1 59
 LIST OF FIGURES
Fig. Title Fage

Graph of Return (cents/lb. fat) vs. Mean !.foisture 14

Content, for selected standard deviation of moisture
values.

2 Schematic diagram of plant layout. 17

3 Schematic diagram of the Contimab MC 30 continuous 18

buttermaking machine.

4 Graph of Temperature (°F) vs. Scale Reading for 34

calibration of Varian Chart recorder.

5 Location sampling scheme used in slice analysis. 45

6 Continuous block srunple results. 45

7a Graph of Moisture Content (~& ) vs. Time (minutes) 50

(24/11 /71) .

7b Graph of Salt Content(%) vs. Time (minutes) (24/11/71) . 50

Sa Noisture Cusura Chart (24/11 /71 ) 51

Sb Salt Cusum Chart (24/11/71) 51

9a Graph of Moisture Content (%) vs. Time ( 11 /3/72) 55

9b Graph of Salt Content (%) vs. Time ( 11 /3/72) 55

1 Oa Moisture Cusum Chart ( 11 /3/72) 56

1 Ob Salt Cusum Chart ( 11 /3/72) 56

11 a Graph of Moisture Content (%) VSo Time (minutes) 5S

(6/4/72)
11 b Graph of Salt Content (%) vs. Time (minutes) (6/4/72) 5S

12 Graph of Moisture Content (%) vs. Salt Content (%) 59

(6/4/72)
13a Moisture Cusum Chart (6/4/72) 60
13b Salt Cusum Chart (6/4/72) 60

14a Graph of Mo isture Content (%) VSo Time (minutes) 63

(7/12/71)
14b Graph of Sa lt Content (%) VSo Time (minutes) (7 /12/71 ) 63

Graph of Moisture Content (%) vso Sa lt Cont ent (%) 64

(7/12/71)

16a Moisture Cusum Chart (7/12/71) 65

16b Salt Cusum Cha rt (7/12/71) 65

17a Frequency Distribution Histogram for Mo is ture Values 70

17b Frequency Distribution HistogrQlll for Sa lt Values 70

18 Graph of Cream Temperature (scal e units) vs. Time 76

(minutes) (7/2/72)

19a Graph of Mo isture Content ( ~ ) vso Time (minut es ) 77

(7/2/72)
19b Graph of Salt Cont ent(%) vs. Time (minutes ) (7/2/72) 77

20a Graph of Cr eam Fat Content (fa ) vso Time (minutes) 85

(7/2/72)
20b Graph of Cream Acidity (% Lactic Acid) vs o 'r ime 85
(minute s) ( 7 /2/72)

21a Apparatus used in measurement of Ribbon Speed 96

21b Apparatus used in measurement of salt slurry 96
injection pressure

22 Graph of Ribbon Speed (pulses/second) vs. Time 98

(minut es) (6/4/72)

23 Ribbon Speed Cusum Chart 99

24a Graph of Moisture Content (%)vs. Time (minutes) 102

( 16/5/72)
24b Graph of Salt Content (%) vs. Time (minutes) ( 16/5/72) 102

25a ~1oisture Cusum Chart ( 1 6/5/72) 103

25b Salt Cusum Chart ( 16/5/72) 103

26 Gr aph of Ribbon Speed (pulses/second) vs . Time 104

(minut es) (16/5/72)

27 Graph of Salt Pump PrGssure (units) vs . Time (minutes) 105

( 16/5/72)

28 Ribbon Speed Cuswn Chart (16/5/72) 106

29 S~lt Pump Pressure Cusum Char t (16/5/72) 107

30a Graph of ~1oisture Content (%) vs. Time (minutes) 111

(22/5/72)
30b Graph of Salt Content (%) vs . Time (minutes) (22/5/72) 111

31 a ?foisturc· Cusum Chart (22/5/72) 11 2

31b Salt Cusum Chart ( 22/5/72) 112

32 Graph of Ribbon Speed (pulses/second) vs. Time 113

(minutes) (22/5/72)

33 Graph of Salt Pump Pressure (units) vs. Time (minutes) 114

(22/5/72)

34 Ribbon Speed Cusum Chart (22/5/72) 11 5

35 Salt Pump Pressure Cusum Chart (22/5/72 ) 11 6

36a Graph of Moisture Content (%) vs . Time (minutes) 120

(26/5/72)
36b Graph of Salt Content (%) vs. Time (minutes) (26/5/72) 120

37a Moisture Cusum Chart ( 26/5/72) 121

37b Salt Cusum Chart (26/5/72) 121
38 Graph of Ribbon Speed (pulses/second) vso Time 122
(minutes) (26 /5/72)

39 Graph of Salt Pu.mp Pressure (units) vso Time (minutes) 123

(26/5/72)

40 Ribbon Speed Cusum Chart (26/5/72) 124

41 Salt Pump Pressure Cusum Chart (26/5/72) 125

42 Correlation - Scatter DiagraLl Examples 1 41

 .L I ST OF TABLES
Page

I SUl.unar y of Churning Factors n...rid thei:.· effect on 25

product componitlon

II Typical Churn Opor~ting Conditions 25

III H.esults for moi:-;turo and s·1l t content for slice 38

samples (28/9/71 )

IV Results for moistur e and salt cont(:)nt for like 39

locations (28/9/71 )

VI Rosul ts for rJo is ture and salt content fo r slice 41

srunplcn (?8/10/71 )

VII Results for !JlOicturo and salt content for like 42

locations (28/10/71)

VIII Summary of result::> for the correlation coefficient 53

between battc:r moisture content and salt content

IX Datn for tho frequency di stribution of 111oisturc and 69

sQlt values taken from four trials
(25/11/71' 7/12/71' 21/12/71' 7/2/72)

x Results for the correlation coefficient between 86

values of cream fat co!ltent and butter moisture
content - a displacement analysis

XI Swnmary of r esul ts for tho correlation coeff icient 87

between but t er moisture cont ent and cream fat content

XII Frequency distribution for differences between duplicate 89

analyses for cream fat content

XIII Summary of results for the correlation coefficient 87

between butter moisture content and cream acidity
XIV Frequency distribution for differences between 89
duplicate analyses for cream acidity

Estimat ed measurom0nt accuracy required for 132

measurement of churn variables

XVI Summary of results for va riab les other than moisture 144
and s <J.lt content
SECTION I I.NTRODUC'rION

Continuous buttermaking has become an important process in the

New Zealand Dairy Industry. From the first experimental trials in 1964
conducted by staff of the New Zealand Dairy Research Inst itut e, contin-
uous manufactu.re of butter by the Fritz process has now reached a
commercial production of over 80,000 tons.

Process control of the buttermaking process pl ays 2.n important

part in determining the economic return on the product. At this stage
adequate information is available on the general oper ating principles
of Continuous churns . However , in the absence of change in the operating
conditions the product continues to show compositional variation. There
is little information available to determine the causes of such varia-
tion. The work undertaken during this proje ct was primarily aimed at
investigation of this variation. It was conducted as a process capabil-
ity study.

It is clear that Continuous Buttermaking will continue to be an

important process in the New Zealand Dairy Industry for some time, and
it is essential that further information regarding the process is made
available in order to facilitate i mproved process control.

An Introduction to Continuous Buttermaking

Continuous manufacturing processes for butter may be divided into

two broad groups.
(1) Churning of cream of normal composition (35-45% fat) such as the
Senn and Fritz processes. The cream is churned in horizontal cylinders
by high speed beaters. The butter granules formed are separated from
the buttermilk and this may be accomplished in several ways. The
granules are washed and then pass into a working section. Salt, in
slurry form, and additional water may be injected into the butter during
the working period. The final product is extruded from the worker and
may then pass to a packing unit.

(2) Reseparation of the cream to give a high fat content cream.

Destabilisation of the emulsion following the second separa:tion may be
used to facilitate concentration of the fat . Accurate quantities of
other constituents such as salt, water and flavouring are added to the
 2.

concentrated cream to obtain a standardised mix of that composition

desired in the butter. The standardisation may be accomplished either
by batch or continuous techniques. The standardised mixture is then
cooled, agitated and worked by various methods to obtain a product similar
to conventiona l butter .

The butter texture obtained by recent modified processes of this

type is satisfactory in comparison with conventional butter and is an
improvement over earlier techniques.

This type of process may also be used to manufacture recombined

butter from anhydrous milk fat, skim milk powder (or serum), salt and
water. Processes of this typo include New Way Process, Alfa, 1i.lf~ Laval,
Meleshin, Cherry Burrell , Creamery Package and Kraft.
 General Process Block Diagrams

(1) Churning cream of normal co~position.

Raw Milk --+ Separation Cream _. Cream Cooling

I Treatment Fat Crystallisation

(2) Reseparation Methods.

(a ) Manufacture from hi gh-fat cream .

Separ ation .....

l Cr eam
Treatment
Reseparat ion

..... Standardisa ti on ..... Cooling

Agitation

(b) Manufacture using s epar ate.d milk fat.

Separati on Cream Re separation .....

Treatment

G
l'hase
version
--+ Removal
of fat
·----~
St andardisati on ..... Chilling
Agitation
 Continuous Buttermaking in New Zealand

Continuous Butter mnnuf3cturc in New Zealand has centred around

use of the Fritz process which involves churning crc.::u;:i of 35fo-45% f at
content. This process was f a voured over the 'reseparation ' method for
s ever al r easons .
(a ) The product was very similar to tha t of conventiona l churning.
(b) It was likely that the differ ent t exture of butter manufactured
from the r e s epar a tion met hods would not be ncc eptnb l e t o consumers used
to conventional butter,
(c) Plant costs and l ayout were less favourable for the r es epar a tion
method.

Recents developments, however , have r esulted in a change in these

f a ctors . A modified r esepar ation method may now become acceptable when
these factors are r eviewed.

An Alfa continuous plant was used t o manufacture butter in New

Zea land in 1947. This WQS the earliest commercial t ria l of Continuous
butter cianufacturc. This pr ocess was r e j e cted due to difficulty in
moisture control and unac ceptability of physical cha r a cteristics.

A Contimab }ffi5 continuous butt er mnking machine was selected for

trials by the Da iry Resea rch Institute. It was inst alled during October
1964 in a fact ory of the Manawatu Cooperative Dairy Company and operated
by staff of the Dairy Research Institute. La t er, the machine was
purchased by the Manawatu Cooper a tive Dairy Company and used to produce
butter , for patting. to be supplied to the local market. This machine ,
using the Fritz process, wa s able to manufacture selted butter. Previous
machines had been unable to flnnufacture salted butt er and as a ll but a
small proportion of New Zeal and produced butter i s s alt ed, these earl i er
machines were not suitable for New Zealand conditi ons.

A second Fritz process machine, a Westfalia BUC 1500 was l a t er

i mported for trial by the New Zealand Cooperative Dairy Company and f irst
operated in the autumn of 1966. After modifications to the working
section and salt s lurry injection system this continuous churn was in
regul ar operation during the summer of 1966/67 .

At the present time the New Zealand dair y industry has more than
twenty continuous butter churns. Three types of machine are in use in
New Zealand,
Contimab (France)
Silkcborg (Denmark)
Westfalia (Gernany)
In each case the Fritz process is the basis for production of butter.

A considerable ar:J.ount of information has been compiled during

investigations and trials conducted on continuous buttermaking onchines
in New Zealnnd . Much of this has been the result of work carried out by
staff of the Dairy Research Institute. This information deals with the
gener al operating technique of continuous butter oaking, covering cream
composition and treatnent, churn operntion and butter packing. These
oper ating instructions provide a guide for manufacture of quality butter
throughout the dairying season, during which considerable changes in rau
material and other factors may occur - Fron this information an a cce ptable
product which meets compositional, microbiological n..nd textural r equire-
ments may be produced.

It is well kno1m by butter na.kers, that even with correct control

sGttings , a good quality r nw material and .~sonag~:y__steady conditions
the butter produced by the machine hns n significantly va rying compos -
ition. The control of butter quality par arueters can be a complex task .
Variations in product character istics m2y occur for no ' apparent ' r eason ;
e.g. when no changes hove been made in operating conditions. Such changes
nay be caused by a l arge number of factors, many of 1·1hich the operator
is e ither unaware of, or unable to control.

The object of tho project m1s to obtain information concerning the

behaviour of the butter manufacturing systera. Tho important parameters
of performance must include those on which the product is graded and
which deter mine the e conoLlic return.

The objective characteristics of moistur e and salt content are

important factors in determining monetary r eturn - as high values for
moisture and s.al t content will result in favourable return, yet exceeding
the maximum permissible limit will incur monetary penalties and in
 6.

ext r eme cases rej e ction of the product. Modification of rej ect ed product
to enabl e acceptance will incur further cost.

The butter churn operator is mainly concerned with moisture and

salt l evel s in his product? as the product quality and his own perform-
ance ar e judged by these factors. Microbio l ogical s tandards pl ay .'.lil

important role in butter quality - but these f ac tors are not available
fo r examination as readily as the o oistur e and salt va lue s.

From the analysed Moistur e and Salt r esult s the operator behaves
as a 1
f eedback 1 link to t he machine . He a cts to control the machine
output by suitable operation of control oechanisms.

The object of the r esearch was to:

(i) measure and r ecord variation in butt er moisture and s a lt con tent
as a study of process capab ility
(ii) assign some causes to the variations where such causes were
suggested by the r esult s of the i nvestigations.
It was not the purpose of this work to define exhaustively the ass i gnable
ca uses of varia tion in this process. If suff ici ent inf oITJ.ation could be
obta ined it mi ght be possible to improve cont r ol of butter nanufa cture .
Reduction in corapositi onal variation will enab l e changes to be made in
manufa cturing liEJ.it s and it i s like l y an increase in mone t a ry r eturn
will r esul t .

Throughout the investigation a tt ent ion was paid mo r e to 1

macro-
varia tions 1 and trends in compositional change . Experimental methods
wer e selected conside ring several factors. The ac cura cy requir ed in
compositiona l de t ermi nation was a ma j or consideration. Highly a ccurate
analytical r esults r equiring a considerabl e amount of time and a ttention
to detail were not necessary to determine the significant trends and
changes occurring in butter composit ion.
 Literature Review

At the present time information regarding the process of continuous

buttermaking is generally related to determining process conditions
which can result in manufacture of an acceptable producto Since the
composition of an acceptable product has relatively wide tolerances,
such information deals with 'coarse' control of process variables.
Si.gr1ificant changes in product composition may be corrected by approp-
riate changes in processing conditi ons and churn operation of this type
has been documented for New Zealand conditionso

Dolby, J eb son, Le Heron ( 1 965) ( 1 ) of the New Zealand Dairy

Research Institute have discussed the ope ration of a Contimab MB 5
continuous churn during a trial of six monthso A later report by
Dolby (1967) (2) considered operation of a Westfalia BUC 1500 and
similarly the emphasis was on achieving control such that a suitable
product was manufacturedo Anderson (1 969) (3) discussed commercial
operation of a Silkeborg continuous buttermaker over a period of
eight months.

A more detailed review of these investigat ions and their results,

is presented in Section I in a discussion of continuous butter manufact-
ure.

Little or no information has been available regarding patterns

and trends in butter composition during manufacture, when the churn
has been running under production conditions with infrequent control
changes.

Determination of Butter Composition

The composition of butter has an important effect on the economic

return achieved for the product. It is in the best interests of the
buttermaker to maintain a high overrun yet not exceed legal requirements,
or offend the consumer.

Because of this, considerable effort has been made to facilitate

control of buttermaking. In particular, research into continuous
measurement techniques for evaluation of butter moisture content has
 80

been extensive. It is hoped that accurate continuous measurement of

butter moisture content would enable feedback of information to some
control system which would act to correct fluctuations in compos ition
by adjustment of process factors.

Two techniques have been investigated extensively.

(i) continuous measurement of the dielectric constant of butter
(ii) continuous measurement of microwave absorption.
The quantity of microwave ener gy absorbed by a material such as butter
is related to the moisture content of the mater ial.

Pa.rkash and Armstrong (1969) (4 , 5) investigated the dielectric

constant for butter and found that several factors contributed to this
parameter. In a particular butter, increasing moisture or salt
content tended to result in an increased dielectric constant; and
in butters worked for the same time it · incr eased aL~ost : linearly
with moisture content. However, the dielectric constant decreased
markedly with increased working time until a steady base value was
r eached. This base value was considered to be a function of the butter
moisture content. For a butter of constant moisture content, the
dielectric constant increased almost linearly with salt content .

Thus it is seen, that under production conditions where

variation in working time could occur, fluctuations in salt content
and also temperature will tend to vary the dielectric constant - as
well as the measured parameter - moisture content.

Sone, Taneya , Randa (1970) (6) discussed conditions for measure-

ment of dielectric constant in continuous butter manufacture.
Using a frequency range from 300 KHz to 1 MHz it was found to be
possible to detect changes of as low as 0. 1 %moisture as a r elated
variation in dielectric constant. Temperature correction was
considered to be necessary to a chieve greater accuracy, as the dielectric
loss factor is dependent upon temperature in the frequency range used.

Koenen (1968) (7) describes an instrument developed by Brabender

for continuous measurement of dielectric constant of butter. The
dielectric constant is then related to a moisture content and a
digital or graphical r eadout is given.
 9.

This instrument is claimed to have an accuracy of better than

.± 0. 2 %moisture i f run under suitable conditions. These conditions
incl ude :

(a) a discrete distr ibution of moisture droplets , also indicative

of adequate working.

(b) stable inputs of cream composition, flow rate, machine beater

and worker speeds.

(c) constant back pressure in the extruded ribbon , this may be

influenced by behaviour of the packing unit .

(d) constant butter density - a change in air content of 1 . 0 %

may gi ve rise to an indicated change of 0. 1 %moisture.

This instrument may also be used for analysis of salted

butter.

Fexa and Rosenbaum (1965) (8) describe a Czechoslovak Akvameter,

a moisture measuring device based on mea surement of electrical
permittivity which is r elated to dielectric constant. The claimed
accuracy was .± 0. 3 %moisture, with a sensitivity of ± 0. 05 %moisture.

Cerna and Vedlich ( 1965) ( 9) continued analysis of the perfor m-

ance of the Akvameter. It was found necessary that moisture droplet
diame ter should be l ess than 20 microns. Calior ution of the instrument
was used to compensate for pH, solids-not-fat and air content variation.

Fexa and Rosenbaum (1966) (10) and Fexa (1967) ( 11) discussed
automatic regulation of butter moisture content by use of the
Akvameter in conjunction with a Cz echoslovak continuous butterma.king
machine. A control system is proposed.

It was reported by Doleza lek and Rosenbaum (1970) (12) that i n

use of the Akvameter MA- 2 , of 34 observations most results showed
close agreement with those obtained by standard moisture tests.
Four r esults differ ed by ± 0.2 %moisture, fourteen by± 0. 1 %moisture
and the remainder were in exact agreement.
 1 o.

It was said to be importnnt that manufacturing conditions were

stabilised; water dosing was not used as if there was insufficient
working after dosing the free moisture r esultant in the butter had
a deletcr.ious effect on the a ccuracy of the meter.

The long term ac curacy under the variable conditions like ly to

be found in normal manufacture was not discussed.

The National Research Council of Canada (13) has developed a

microwave instrument for measurement of water in butter and an a ccuracy
of two parts in ten thousand in determination of butter moisture is
claimed. It is not stated whether thi s applies to both salted and
unsalted butter. However, r ecent information indicates that work on
development of this instrument has not been successful and the project
has been terminated .

Slight (1970) (14) describes a similar microwave technique but

indicates an accuracy of up to.± 0. 1 %moisture , depending also upon
the material and its method of presentation. No specific details Qre
given for anal ys is of butter o
 11.

Conclusions

Two basic types of continuous moisture measuring instruments

have been deve lopedo At present neither instrument has been fully
tested and accepted for New Zealand conditionso

Variability of raw materials , machine factors and butter compos-

itional factors such as salt and air content reduce the reliability of
the instrumentso However, if a suitable instrument is developed, it
is likely that it will become part of a more complex system designed
to control butter moisture content.

Little emphasis has been given to continuous salt estimation,

but since this is an important economic factor with respect to overrun
it is likely that continuous salt measurement in butter will be a
further requirement for any control system utilising error detection
in the output, with information feed back to the processing unite
 12.
The Economic Importance of Butter Composition

The monetary return on butter is rela ted to its composition and the
overrun achieved. Moi s ture, salt and curd content determine the overrun
and the return per pound of fat may be ca lculated from the basic price
paid per pound of butt er and the known compositional values.

R = p
1 - (rvr + s + c)
where R - return (c ents/lb. f a t)
p - price for butter (cents/lb.)

M - Fractiona l Moisture Content

s - Fra ctiona l Salt Content
c - Fractional Curd Cont ent

As the moisture content increases the return per pound f a t will

also increase, but the no isture content is limited to a maximus. of 16.o%.
For values above this limit, a penalty is imposed. Assuming the m·oisture
va lues to approximate to a ' Normal Distribution', it is possible to
det ermine the quantity of product which will fall both inside and outside
the tolerance and hence the r eturn on the product.

If it a ssumed that salt and curd values a re constant, then for a

s elected standard deviation of the mo isture va lues , the return on the
product may be calculated for each overall moisture nean value.

Assumptions: Price/lb. butt er p

26.26 cents
Penalty for overmoist butter , a = 2 cents/lb.
Salt Content = 1 .6 %
Curd Content = 1 .1 %
R P - aK
1 - (M + S + C)
where K =fraction of produce exceeding 16.o% cioisture, determined
from a Normal Distribution of the stated standard
deviation.
M = average moisture content of butter.
 13 .
Sample Calculation
Mean moisture = 15.9 %
Standard deviation of moisture values, S = 0.1 (selected value)
Transformation to unity Normal Z 16 M = 0.1
Distribution , s 0.1

= 1.0
Area to right of Zin Normal Distribution, K 0.156

1 + s + c (15.9 + 1.6 + 1 .1) %

= 18.6 %
= 0.186

R = P - aK cents/
1 - (M + S + C) lb. fat
= 26 .26 2(0.156) cents/
1 - o.186 lb. fat

Return 34.34 cents/lb. fat

For a series of sel e cted standard deviations the return was

calculated for various mean moisture values.

Fig. 1 (page 14) is a graph of the Return (cents/lb. fat) vs" Mean
Moisture Content for selected standard deviations of the moisture values.

Discussion
It is seen from this graph that an optimal return may be achieved
for each standard deviation of the moisture variable by selecting the
correct mean moisture, the greatest return being realised for the
lowest standard deviation. The practical situation does not allow such
precision , but these calculations can provide a suitable and practical
approach to improving monetary return.

A similar graph may be drawn to indicate the varying economic

return if other compositional factors such as salt content and curd
content are treated in the same manner.
 14

Fig. 1. Graph of Return (cents/lb. fat) vs Mean Moisture Content

for Selected Standard Deviation of Moisture Values
Return (cents/lb)
35.0

34.0

L----i___.___..__-'---~--......~~---------~
33.015.0 15.5 16.0
Moisture (%)

MASSEY UNIVERSITY
 THE CONTIJVIJill CONTINUOUS BUTTERMAKING MACHINE

Churning
A variable speed mono pump supplies cream of fat content 38-42%
to the single stage horizontal churning cylinder, The cream enters
tangentially to the four bladed beater rot ating in the churning cylinder.
The beater speed can be varied (by means of a variable-diameter, expand-
ing pul ley) from 1000 to nearly 2000 r.p.m o

l!"'or churning of low fat creafi or of cream having 'ha rd' fat,
increased turbulence in the churn raay be achieved by insertion of
po rfora.bod lin 2 r R in th0 r.hurni. ng cyJin c10r. Chill ed w11.tP.r m.<1.y be
circulated in a jacket surr ounding tho primary churn.

The butt e ~ granules and butt ermilk pas s from the churning section
i nto the base of the working section. The working tunnel is inclined to
allow drainage of wash water and buttermilk from the various sections.

Working Sectto11
The butter granules flo a t on a pool of buttermilk in the base of
the v:orking secticn. Twin r otating screws carry the butter granules
forward to the first operat ion which is buttermilk drainage .

The buttermilk drains from the working section through a sieve

which retains any granules flushed from the worke r. The level of butter-
milk in the working section is determined by the level of a draining
syphon and may be vari ed in order to a lter the moisture content of the
butter.

The granules are carried by the s crews into the butter drying
tunnel where buttermilk drainage occurs. A variable profile pressure
plate restricts the butter outflow. Reduction of the exit profile
increases the back pressure in this section and assists in buttermilk
removal. This pressure plate is able to be used to vary the moisture
content of the product.

The butter is extruded in thin bars and passes into the washing
section where it is sprayed with chilled water and then is forced through
orifice plates after which cruciform beaters act to mix the butter. The
wash water drains through a sieve and is taken from the working section.
 1 6.

Salting of the butter i s the next operation. A salt-water slurry

is held in a well agitated t ank. The slurry is injected by a diaphragm
pump into the working section through three points located around the
working tunnel . The butter, which is in ' spaghetti ' form when the salt
is inject ed, undergoes further working and mixing as it pa sses through
orifice plates followed by cruciforr:i beat ers. The s a lt inj e ction vulume
may be controlled by varying the stroke of the di aphragn pump.

The butt er passes through a va cuum chamber and then into the
final working section which again consists of orifice pla t es followed
by cruciform be at ers. The butt er i s extruded fr om the outlet head into
the packing unit.

The size of perfora tion i~ the orifice plates can affect the
amount of working thu butter r e ceives . Sna lle r dinneter perforations
uaybe used to increase the amount of working . Sinilnrly, r estriction
of the outlet he2d size i ncr eas es the a Jount of working the butter
r e ceives, as the back pressure in the working section is increased.

Fig. 2 (page 17) shows a general diagram of plant l ayout.

Fig. 3 (page 18) shows a schematic diagr am of a Contima b continuous

buttermaking machine.
 17

Fig 2 SCHEMATIC DIAGRAM OF PLANT LAYOUT

-
- 20 f
-1
;-Sleet

Cream Storage Silo Balance Cream Pump Contimab MC30 Continuous Salt Injection
Tank Churn Pump
 18
Fig. 3. Schematic Diagram of the Contimab MC 30 r l

x
~
Prima ry Churn (Beater)
j :~
l
~
II ...-, -
_t:S>j .
~ .- ~ -
J, -
l
~
J
• •lit
I •Jlr.
In
• .t9't
- r -.l"<-
; ~ ---1.-_ '15.
-···o
. . ... _·=.--~---
Vacuum Section
Water Dosi ng Extrusion
"'- - .1._.} ~

---
1 -~~,~
~
""'-c;__ - - -
- ___ J
- - - -r-_
_J.L

Plate
Wash Water Sprays
-ur1/
r-:;--- - ---- :
fiHl I
-- ·~ --:i
•
- I

,~~ ~u
-:» ~
~ [ --
.!_ . , _

f
~ ~
~ [ _~ ·~..._ 1\ Working Section

~n
·; ;-J \ ',. . .:_ -: -;.
~
• ..i '"·· _,
'
: •• - · '\ \ - 0 ~ ~..1-..
l ,"'"\-'~
•- \. i ,-. ~,, . -
-1 -~ ~ ~ashwater
'"I)

J . ~~'\'•. J·.1_.,:tf·, ~ Ni~ '""- Butte~milk

-~ ~ ~ ~· =-- ~\.-. ~
" , - • • • • -:-,.
-
11
./ : Drain

I
' MRi "' . .
'
I1
I
/o'
r1

,.._ ~ jt', ~~
I
.-1-~
..~~--....::
-
<>-
1•
, ~
•

(
•• \ -
~JTTTITJ
.
,_.-.l l [;~
, .. I -
--i~
.
:
}
~•_L. •
'j
-s ~
~
I
~
\
'
""~~ ·-,
••
.•'-:~--~'.'.''".::.... _
~·K
......
,,
.:.-,_,
Dram

-
:
~ i ~
Yl 1 r ""-;;-"',.
Hit]'--~ - , .
~~
~
- -
~
--

· '=-;c,--1~:E - ,~ ~
~·.~~
I .

-· ~,,-
.. 1·:
-; . r= _.:._ . ..._.._-.=
.. . r~ ~ .l:
1~~- !=._;,'hr~-4r£o~fi~ ~t:- --~~~-rw-1 1°
::.~ 11- -"~''-'-"'-'.,_= -~
~Jilil1ll r.J i1
I ·

'~.!~~- ~-=--
_. · "
'f :,-,_..!.,.
•
-' _,
t· ,
1'
, .-1
I
t _j'O
J n--'!B\ _ ~..,...j
Lii~-
.
- -1 8[61
-~
o · '

~~~~ _ ~-'.#'N"'..rt'"A~it,'L~ -- -.!o ,.,..._,£::!=~- ...J=-...c ~ ffi_ ~

<'<', ....
I I
1
~CT -- - - <!..! -tf: -- ~
~ rA :'_-~ 1~:=io
-

\
--·
] - - - --- .

\Jr ~ -,1;1-
.

~ -~
.....
9I
1
L ~,'~ ~~ o~~c::; ~~ ~
= =
C::J
.QQ. 00

._1~
o=o=o
J:;::::JC
00

t=J::::::::>
0000

c=x=:it=1~
[]L]r:::-i
= = ~
00 00 00
o7o c:::n:::=>
c::::::>::::::::> c::::::>::-i~
 1 9.
Factors Affecting Butter Composition

The following discussion is confined to the operation of the

Contimab continuous churn. In some instances quantitative effects are
considered and these are eiven only as a general guide. Interactions
of several factors may alter conditions sufficiently so as to change
the quantitative effects quite considerably. Similarly, extreme
conditions of operntion may cause unusual behaviour.

Churning of the Cream

Generally, variation of a cream or machine character istic 1·1hich

tends to facilitate churning of the cream tends to increase the moisture
content of the butter.

Cream Characteristics

(a) Fat Content

The fat content of the cream may be standardised to an optimum
value of 40-41 :· ~ for continuous buttermaking.

High fat cream (45/~ fat), al though assisting granule formation,

may be difficult to churn because of its high viscosity. A low fat
cream (35~ ) is difficult to churn because of the high proportion of
serum to fat. This hinders fat globule agglomer~tion.

Vari ations in fat content may give rise to changes in the moisture
content of butter . It has been shown that an increase in cream fat
content of 1% may give rise to a butter moisture content increase of
o.6% if no compensation is made by alt eration of other factors. (1)

(b) Butterfat Characteristics

The butterfat composition is important in determining the type of
cream treatment used prior to butter manufacture. The treatment given
is able to influence the churning character of the cream. It also
influences the fat properties which provide the textural character of
the butter.

A soft fat is one which has low softening temperature. Such a

fat may be plastic even at sub-ambient temperatures.
 20.

A hard fat is one which has higher softening temperature and it

may be hard or solid even at temperatures exceeding ambient. Butter
fat is a mixture of glycerides having different softening temperatures.
The texture of the product is determined by the composite character of
the mixture.

Milk produced in early spring generally has soft butter fat. The
hardness of the fat increases during the spring and summer until it
reaches a maximum in December . A gradual decrease in hardness occurs
during the remainder of the season.

The spring cream having a soft butter fat is shock cooled after
pasteurisation-vacreation and is generally churned at lower temperatures
0
near 45-48 F. The summer cream having a harder butter fat requires slow
cooling after pasteurisation-vacreation and the churning temperature is
higher being in the region 53-55°F.

Fat globule siz e may also affect churning of the crear;i. Cream
having large fat globules will churn more readily than crean with small
fat globuleso The globule size may be affected by the type of milk
produced by a particular breed of cow and also by cream treatment.

Excessive pumping and agitation tends to break up fat globules,

especially when the cream is warm. Correct cream handling may avoid
this. The pasteurisation operation may influence fat globule size also.
Indirect heating such as in a plate heat exchanger a ssists in fat
globule agglomeration. Direct steam injection ha s u reverse effect and
disruption of the globules may occur. Variations in the conditions
during this treatment may result in a variable raw material for process-
ing and this could cause variation in the product .

(c) Cream Acidity

The acidity of the cream can influence its churning ability.
A creno having acidity higher than normal tends to churn uore
readily - and moisture content may rise if compensation is not made.

In the Contimab machine, when cream acidity exceeded 0.13%,

difficulty was found in maintaining the moisture level below 16.o%. (1)
For high acid cream, such as this, variation in other controlling
 21.

factors is necessary such as lowering of churning temperature or

decrease in primary churning by lowering of beat er speeds .

Varia tions in crerun acidity will thus tend to give rise to

variation in the moisture cont ent of the product . It has been shown
that an incre a s e i n cream a cidity (d etermined a s l actic a cid) of 0.01%
can give rise to an incre ase of 0.5% in butt er moisture cont ent. (1)

(d) Cream Churning Temperature

This characteristic is important in continuous buttermaking.
The churning temperature is to some extent determined by the crean
tr eatment. This i s a function of the type of cooling given to the
cream as it come s fro m the treatment unit and also of the temperature
cycle mainta ined during overnight storage in the holding silos. An
increase in cream churning temperature generally r esults in an increase
in mo isture cont ent of the butt er as the cream churns more r eadily. The
converse is also true .

Sone measure of cont r ol may be exer t ed during manufactur e by

using the chill water j a cket on the primary churn. Temper ature changes
of 1-2 °F. may be a chieved a lthough t his change in tempera ture does not
have as great an eff e ct on the butt er noisture content a s would a similar
change in temperature of the bulk cream. It has been shown that an
increase of 1°F. in bulk cream temperature can give rise to a butter
moisture cont ent incre ase of o. 5%· (1 }

Machine Vari ables

(a) Primary Churn

The primary churn bea t er speed is a prime source of control for
moisture content of the product. By increasing the bea ter speed the
amount of churning of the cream is increased and the butter moisture
content rises.

Experimental work has shown that an increase of 100 r.p.m. in

beater speed may result in a butter moisture content increase of up to
0.5%, but the relative effect varies considerably with changing condi-
tions. ( 1)
 Control of beater speed is by a variable-diameter expanding
pulley activated through a servo Bechanism.

The beater speed ranges from 1000 - 1800 r.p.m. on the Contimab.
Excessive speeds for churning of low test cream can result in "thrashing"
of the cream and consequent excessive fat losses in the buttermilk due to
fat globule break up. Soft spring fat may require beater speeds as low
as 1000 r.p.m. At this level, subject to cream flow, the motor may draw
a current of only 40 Amps. Harder summer fat may require a beater speed
of up to 1800 r.p.m., at which the motor may draw 72 .Amps .

The range of beater speeds used is dependent on the type of cream.

However, speed variation about this ' base ' speed may be used to control
moisture.

(b) Cream Flow Rate

The rate of cream flow to the primary churn is another controllable
factor. Reduction in cream flow rate over a limited r ange can allow a
greater amount of churning and so causes an increase in moisture content.
Likewise an increase in cream flow rate can reduce butter moisture
content.

Changes in cream flow rate in order to change butter output

generally result in a change in the butter moisture content and so
necessitate process adjustments.

(c) Working Section

The working section of the Contimab may be used to control moisture
and salt levels in butter .

(i) Speed of worker screws

The important source of fine control is the speed of the worker
screws. Increasing the speed of the working screws causes an increased
quantity of buttermilk to be carried forward, thus incorporating it
into the butter. A decrease in worker speed will achieve a decrease in
moisture content. .An excessively low worker speed can result in under-
working of the butter with consequent poor texture and moisture distrib-
ution. The maximum useful worker speed is generally limited by the
moisture content which tends to increase with worker speed, and also the
 23.

capabilities of the driving mechanism.

The worker speed may range from 30 r.p.m. with spring cream (soft
f a t) to 60 r.p.m. with summer cream (hard fat).

(ii) Level of buttermilk

The level of butt ermilk in the base of the working section may be
adjusted by varying the height of the gooseneck syphon which drains the
buttermilk . A higher level of buttermilk will tend to incre ase the
quantity of buttermilk carried forward and thus increase the butter
moisture content. This is a rela tively coarse control and is not
generally varied during a run. Better control may be achieved by
changing other f a ctors .

(iii) Restriction plate s in the working section

Restriction plates within the working s e ction can be a lt ered in
order to change the butter mo isture cont ent . The restriction at the
butt er drying section is significant. A r educed area (incr ease in
r e striction) will caus e the butt er moisture cont ent to drop.

The r es triction pla t e s also control the amount of working the

butter receives, particularly in the l a tter section a fter salt slurry
inj e ction.

(iv) Dosing pump

A f eature of the Contimab MC 30 ma chine i s the 'moisture dosing
pump'. This diaphragm pump is used to alter the mo isture l evel in the
butter. It is situated prior to the final working section, just after
the vacuum chamber and it injects water into the butter. By varying the
flow rate of this pump the moisture level may be varied. It is more
convenient to alter the setting of this pump than to change another
factor which may give rise to other changes in interrelated factors .
The dosing pump is thus an 'independent' unit for fine moisture control
and may be used to add up to 0.15% moisture into the butter.

(v) Chilled buttermilk recycling

Recycling of chilled buttermilk was used on the Contimab MC 30
studied during the trials. The recycled buttermilk sprayed on to the
granules chilled them, thus firming the granules. This may have an
 24.

effect on the moisture content but the temperature and flow of butter-
milk are generally not used to control moisture content.

(vi) Wash Water

The wash water temperature can also influence moisture content.
However, it is not a convenient method of control.

(vii) Salt slurry in.iection

Salt is incorporated into the butter by injection of a salt/water
slurry. The composition of the slurry may be varied, but a slurry of
60 parts salt/40 parts water by weight was used in the machine during
experimental trials.

The salt content of the butter nay be varied by control of the

slurry injection. This is achieved by changing the dosing volune.
Because of the water associated with the salt a chruJ.ge in dosing volume
in order to va:ry the salt content will r esult in a corresponding change
in the quantity of water injected into the butter .

(viii) Angle of inclina tion of the working tunnel

An increase in the angle of inclination of the workine tunnel
tends to result in improved buttermilk drainage and so r educes the
moisture level. This is not normally changed during a run, but the
angle may be increased to promote drainage during periods of difficult
moisture control e.g. with spring cream having soft butterfat.
 25 .
Operati on of Contimab Continuous Buttermakcr

The preceeding discussion h~s indicated some of the raw material

and nachino variables which determine the composition of the end product.
These factors are interrelated, thus to achieve a change in composition
an alteration nay be made to one or more of these control factors.

A sUDBary of these control factors and their effect on composition

is presented below in Table I.

TABLE I Moisture §ill.

Factor Change Content Content
Fat Content of Cream Increase Increase
Acidity of Cream Increase Increase
Cr~Ll~ Temperature Increase Increase
Churn Beater Speed Increase Increase
Cream Flow Rate Increase Decrease
Chill Water Temperature Increase Increase
on Prinary Churn ( SI!lall)

Worker Speed Increase Increase

Butteroilk Level Rise in Syphon Level Increase
Uorker Angle Increase Decrease
Wash Water Teoperature Temper ature Increase Increase
Salt Puop Stroke Increa se Increase Incr ease

TABLE II An Indication of Typical Operating Conditions

Spring Cream SULlBer Cream

(soft fat) (hard fat)
Fat Content 40.5 % 40. 5 %
Churning Temperature 45°F. 52°F.
Beat er Speed 1000 r.p. m. 1800 r.p.m.
Amperage 40 Araps 74 Amps
Worker Speed 35 r.p. m. 55 r.p.m.

Note 1 - A list of references used in preparation of this section is

pr esented in Section B of the bi bliography.
SECTION II. METHODS OF ANALYSIS

The f a ctors selected for ana l ysis were chosen primarily because of
their direct bearing on the economic return received for the producto
Such factors as moisture and salt content uhich influence the overrun
achieved on the product a r e in thi s categor ya Other factors which
affect grading of the product may also be included o

In additi on , those f a ctors which influence process operation,

having an inclirect effect on the economic i~eturn, have been considereda
For example , butter moisture content is dependent upon a number of cream
and machine operating f a ctorsa Similarly, butter salt content is also
det ermined by sever al machine operating factorso

The me t hods used in analysis of the factors were required to

s atisf y certain conditions.
(1) The method must be comparable to that used in the factory labora-
tory , as it i s by the r esul ts obtained f rom the laborat ory tha t the
operator controls the continuous churn. As i t is the variation in churn
operat i on being exami ned, highl y accur a t e D.nalys is of the pr oduct is not· ·
necoss3.r;-/ uhcm the pro co s3 is controlled to l esse r precision.
(2) Similarly, the method should be comparable to t hat used in gr ading
of the produc t uhich influences product r emune r at ion a
(3) The method shoul d be uncomplicated and suit ab l e for examination
of many samples "
(4 ) The accuracy of the me thod should be sufficient t o detect changes
and trends i n t he factoro Hi ghly sens itive techniques are not required a

Sw:nplin,g: Errors
Sampling e rrors are introduc ed into r esults when non-homogeneous
products such as butter or cream ar e sampled , Moisture droplet dispersion,
fat globule distribution may be homogeneous under ideal mixing conditions
but under pra ctical operahng conditions this may not be trueo

The extent of non-homo geneity is difficult to assess as the

differences involved can become incorporated with experimental errorsa
Thus differences between duplicat e or multiple samples reflect a
combination of errors in sampling and experimental technique .

The use of larger samples can reduce sampling errors to some

 27.

extent, as localised minor variations will have a less significant

influence on the result.

Throughout experimental work samples of more than 8 grams weight

have been used. (Further discussion of errors is made with reference
to individual tests.)

Determination of Moisture in Butter

The method selected for moisture analysis of butter was basically

that method used in the factory ' Moisture Test Room ' . It is also
generally used by Dairy Board Gr~ding Officers and thus is the basis
for monetary return on the produce.

The method is not difficult to use and a satisfactory rate of

analyses can be achieved.

Modifications 1rnre mo.de to the cooling procedure for dishes after

drying when empty, and after heating of the sample. This was done in
order to avoid introduction of errors due to weight changes when
cooling. Adoption of a standardised procedure assists in elimina tion
of errors.

All weiehing measurements were made to a four decimal place, s ix

significant figure accuracy using a Mettler H6 balance . Thu final ~ esults

were expr essed to two decimal place , four significant figure, accuracy.
Duplicate analyses were made from each sample and results were finally
expr essed as the mean of the duplicate va lues .

Summary of Procedure for Butter Moisture Determination (26)

Apparatus Moisture beaker

Moisture tongs
Flame source, hotplate
Balance
Sampling 1 trier'
Method (1) Dry beakers in oven overnight (1oo 0 c. )
Remove and cool for a standard time of 1.5 hours on the
balance room bench.
 280

(2) Weigh all dishes, set one a side as a standard (zero balance
before use).
(3) Weigh accurately a butter sample of 8-10 grams.
(4) Preheat sample on hotplate, complet e moisture removal by
heating over flame till curd becomes a light brown colouro
(Take care to avoid splashing of fat from dish.)
(5) Cool for standard time 2.5 hours.
(6) Reweigh (zero balance before use ) .
(7) Adjust final weights by difference in weight of standard
dish.
(8) Calcula te moisture content.

Analysis of Errors
Sources of error in the experimenta l method are considered in
the following discussiono Assignable errors will be evaluated where
possible and their nett e ffect on the r esult calculated.

Sources of e rror .

(1) Correct use of the balance was important to avoid errors. The
balance was zeroed prior to weighing and care taken to avoid draughts .
(2) Weighing of the dish, dish and sample.
(3) Unde rheating of t he curd , which can l ead to moisture r emaining in
the sample o
(4) Overheating of t he curd , which can lead t o curd decomposition .
(5) Frothing and spitting can caus e f at loss. This may be avoided by
careful control when heating the sample.
(6) Change in atmospheric conditions can lea d to variation in the
amount of adsorbed moisture on the dishes - this can cause weight change.
This source of error was minimised by use of a standard moisture dish
which was not used in the test. Weight change s in this dish occurring
during the period of analysis were determined and adjustment made to the
final weights of experimental dishes.
(7) Non-homogeneity of sample (as discussed earlier).
 29.

Evaluation of Errors in We i ghing - A Typical Cal culation

The er ror possible i n reading the bal ance was assumed to be 1
divi sion of the most sensitive scale .
Scale r eading error = ± 0. 0001 gm.
Consider a typical dish weight = 23.0000 ± 0. 0001 gm.
Dish + sample wei ght = 31.0000 ± 0. 0001 gm.
• Sample weight s. oooo ± 0.0002 gm.
=
= s. oooo gm. + 0. 003%

Dish + sample weight = 31.0000 ± 0 . 0001 gm.

= 29.7600 ± 0. 0001
.• •Final weight
Weight loss = 1 • 2400 ± o. 0002
gm.
gm.

= 1.2400 gm. ± 0. 02%

Fractional Moisture Content = change in samEle weight

initial nample weight

= 1 .2400 gm. + 0. 02%

s.oooo gm . ± o. 003%
= 0 . 1550 ± 0. 02%

= (0. 1550 ± 0. 00003)

ffoisture Content = (15 . 50 ± 0 . 003) %

Thus it may be seen that the assignabl e weighing errors,are only a

minor source of err or. For a l ower sample weight the error will increase .

Since the r esults are considered with r espect to larger variations

i n composition and t r end changes, this source of error is not significant.
Compositi onal r esults were cal culated to three decimal pl aces and rounded
to two decimal places.
 30.
Determination of Salt in Butter

The Acetone Test for Salt in Butter was the method selected for
analysis of samples. It is based on the Mohr titration . Since both
butt erfat and water are miscible \·Tith acetone it provides a medium
whereby the salt present in the butter may disperse readily. The salt
is then available for titration with silver nitrate using potassium
chromate indicator o

For the purposes of this test, the complete sample remaining after
moisture determination was analysed by this method. It was hoped that by
analysis of a larger sample of 8-10 grams, original weight, without
transfer of material ~ sampling errors could be reduced.

Summary of Procedure for Butter Salt Determination (27 , 28)

(1) Warm the sample to melt the butterfat.
(2) Add 30 ml acetone and stir we ll.
(3)Add app~oximately 100 ml of distilled water .
(4) If butter is acid, add a small amount of pure Caco to
3
neutralise the solution. The chromate indicator gi ves
satisfactory r esults only in neutral and alkaline solutiono
(5) Titrate using AgN0 solution of known concentration and 3-4
3
dro ps of K cro indicator should be added.
2 4
Note 2
(1) The commercial acetone used was che cked and fotuld to be free
of chloride.
(2) Distilled water was used in all analyses.

Analysis of Errors
The end-point in t itration of the chloride ion is a possible source
of experimental error. This is due to the cloudy nature of the stirred
solution caused by the silver chloride precipitate. However, the titration
was taken to a consistent end point colour. Because the variation in
salt content is more important than the actual value, having a consistent
end point colour is the major consideration.
Consider a titration error of± 0.1 ml. (2 drops)
Minimum titre values were approximately 7 ml.
Titration error = ± 0.1 X 1 00 %
7
= + 1.5 %of the titre value
 31.

For a salt content of 1.5% actual error is± 0.022 expressed as

percentage salt.

Titre values were generally in the region of 10 ml. and so this

error is reduced =

Analysis of Differences between Duplic ate Results

Differences between duplicate results of a single sample may be

attributed to two factors.
(1) An actual difference between the sample compositions occurring
because of the non-homogeneity of the substance being analysed.
(2) A difference which is a result of inherent errors in the method of
analysis.

An analysis of differences between duplicate results of the tested

samples was carried out . The differences were due to the combined effe cts
of the above factors.

The ~ifferences ' for n samples were used in the calculation of

' mean difference', and ' standard deviation of differences'.

Moisture Content
Number of samples, n = 70
Mean difference 0.029 %moisture

Range minimum difference = 0

maximum difference = o.oa %moisture

Salt Content
Number of samples, n = 70
Mean difference = 0.046 %salt

Range minimum difference = 0

maximum difference = 0.11 %salt

Differences in excess of these quoted values were considered

unacceptable and samples were tested a second time.
 32.

Generally the difference between duplicate s amples was low and at

an acceptable level. In examination of results compositional variations
were compared with the possible magnitude of duplicate differences
when considering real or significant variations.

Determination of Fat Cont ent of Cream

The method used was the s tandard Babcock method for fat determin-
ation in cream. (29)

(An examination of errors is giv en in the discussion of experimental

results.)

Determination of Titratable Acidity of Cre am

The Titratable Acidity of cream was deter mined as lactic acid by

the standard method . (30)
(An examination of errors is given i n the discussion of experimental
results.)

Calibration of the Varian Strip Chart Recorder

The calibrat ion curve of the Varian r ecorder is presented in

Fig. 4 (page 3 4).

The calibration was conducted with r efer ence to a Mercury-Glass

thermometer .
Varian Recorder Model G-1 5-1
Scale - 1 millivolt
Thermocouple Copper-Constantan
Reference temperature - Ice point
Thermometer Mercury-Glass (reading to 0.1°C)
(BS593 - A)

The thermometer was fitted with a magnifier to facilitate scale

reading. It was used to measure the 'known temperature' of a well stirred
 33.

water bath.

The calibration curve approximates closely to a stra ight line, and

it is assumed that over the temper atur e ranee investigated, that the
measured temperature i s directly proportional to the Va rian scale r ead-
ing.
 34

Fig. 4. Graph of Temperature (°F) vs Scale Reading for Calibration of

Varian Recorder

70

60

50

40
25 35 45 55 65

units (Varian Scale)

 35.
SECTION III EXPERIMENTAL WORK

EXPERIMENT I
A study of weight changes during cooling of samples in the
gravimetric analysis of moisture content of butter.

The object of this experiraent was to deterr:iine the magnitude of

weight changes occurring during cooling periods. By cooling the dishes
to a constant we ight or to a l evol of oinimum change in weight, an
important source of error nay be eliminated or reduced to an acceptable
level.

Consider a sample of weight 10.00 grams. The weight loss when

moisture is removed will be ben~ e en 1.5 gm. - 1.6 go . for a calculated
moisture 15% - 16%.

Since a weight change of 1 .500 gm . - 15.0~~ then if we wish to

detect changes in the second decimal place of the moisture content value,
express ed as a percentage, the weighing accuracy must be in the region
of 0.001 gm . - 0.01 ;& moisture.

Thus unless correction is made for warm-weighing the final weight

to be used in calculations should be accurate to the third decimal place
or better.

For smaller samples of less than 10 grams, for a moisture content

value accurate to 0.01%, the weighing accuracy should be taken to the
fourth decimal place.

However, the inherent errors involved in the method of moisture

removal will impose a limit on the accuracy necessary in experimental
weighing.

Weight Changes During Cooling of Empty Dishes

The metal dishes used in moisture determination were dried overnight

in a 100°c. oven. After removal from the oven, the dishes were placed on
a metal tray, they were then placed in a balance room on a marble aggregate
bench and allowed to cool. The dishes were weighed at half hour intervals
until changes in weight were minimal.
 36°

Little change in weight was r e corded a ft er 1 .5 hours cooling in a

sam ple of 24 dishes. The raagnitud e of change was ~ ± 0.0004 gm. be tween
1.5 hours and 2 . 0 hours cooling. Thus 1 "5 hours was se lected as a
suitab l e co oling time for empty di shes aft er oven drying •

.'\. simil ar t::'.'::..al was conducted on butt er sampl es under going cooling
af t er moish:i- 2 hacl been driven off during heating . Fr om a sanpl e of 12
dishe s ·che minim:u:n 1rnight ch:mges oc curred aft er a cooling period of 2. 5
hours, a l though wei ght changes after 2 hou.r.s 1vore reduced to an accept able
l evel. Magnitude of change aft er cooling for 2.5 hours rms ~ + 0.0005 gms.

Thus by usinc a consinteYJ.t procedure fo:r cooling of mo i sture dishes

during the g:cavil11etric mo istm·e determi nat ion of butter , an important
source of e:rror has been removed or r educed to a l ow l eve l which will
help to improve the a ccur2cy and r epeatab ility of resul ts.

Duri ng cooling of the IJ.oisture dish and the c.ish including butter
sampl e the final wei ghts were not constant and a soal l error i s introduced
in the r esul to

The magnitude of these weight changes may be used in ca lculat ion

of the possible eTror.
Coo lin& of dish - chanee after 1.5 hours ~ 0.0004 gm .
Cooling of dish + sampl e - change after 2.5 hours ~ 0.0005 gm.
'I'hus possible err or i ntroduced ~ o. 001 gm .

This could give ri se to a discrepancy of up to 0.01 % in the estimation

of butter moisture cont ent. (Refer to earlier ca lcula tion page 35).

Thu s considering the possible errors assignable to weighing

procedure ( page 29) and cooling pr ocedure, a total a ssignable error has
been estimated.
Weighing procedure error + 0.003 %moisture
Cooling procedure error ± 0.010 %moisture
Total assignable error + 0.013 %moisture
pote 3
A summary of numerical results for this experiment is presented
in Appendix I.
 37.

EXPERIMENT II
The initial invest i gat ion aimed at determining the variations in
moisture and salt content in the product from the Contimab MC 30 when
this machine was running under r egular production conditions.

For the first experimental run slices were cut from the extruded
ribbon immedi a tely before entering the hopper of the packing unit. The
sample slices we re wrapped in a moisture proof film in order to minimise
moisture transfer between the butter surface and the atmosphere during
storage. In addition this protected the surface from water splashes and
. ~ther contamination. The slices were 2.5 - 3.5 inches thick, .this
enabled a core sample of approximately 10 grams to be removed as a single
plug using a 'trier'. Sample slices were removed at specific time ·inter-
vals: O, 1, 2, 3, 4, 5, 10, 15, 20, 25, 50, 75, 100, 125 minutes.

Duplicate samples were taken from the slice at nine specific

pesitions making eighteen samples in total . This was done for each· of
the fourteen slices. Moisture and salt analyses were carried out for
each sample. (Fig. 5 (page 45) indicates the sampling pattern used.)

These results were analysed in two ways:

(i) Each slice was considered separately - standard deviation and mean
values were found for moisture and salt content in the eighteen samples.
(ii) Like positions in the series of slices were considered. Thus for
each specific l oc at ion (1 to 9), the mean and standard deviation wore
calculated for the mo istur e and salt values of the time sequenced series.

Table III (page 38) gives the results for moisture and salt content
for the 14 'slice' samples taken at the specified times. The results
show a wide variation in the mean values for each slice, indicating a
large variation with time. The standard deviation for each slice is
relatively low in most cases, indicating a comparatively even composition
across the slice at any one time.

Table IV (page 39) gives results of moisture and salt content for
each of the nine 'like locations' in the series of slices. The mean
values for each location showed a remarkably low range. These values
are close to the grand mean of all results, indicating that no area in
the slice is being consistently starved or enriched with moisture or salt.
 38°

TABLE III
RESULTS FOR INDIVIDUAL SLICES
TRIAL (28/9/71)

TIME MOISTURE CONTENT SALT CONTENT

(minutes) Mean ( ~ ) Standard Mean (%) Standard
Deviation Deviation

0 15.53 0.05 1 .42 0.1 0

15.70 o.os 1.56 0.06
2 15.75 0.05 1 .45 0.04
3 15.66 0.11 1.53 0.03
4 1 5.46 0.05 1.32 0.03
5 1 5.65 0.04 1.46 0.06
10 15.68 0.04 1.47 0.02
15 15.62 0.04 1.42 0.02
20 1 5.46 0.04 1.46 0.03
25 15.62 0.04 1.45 0.02
50 15.72 0.05 1.52 0.04
75 1 5. 86 0.04 1.41 0.03
100 15. 90 0.05 1. 37 0.05
125 15. 38 0.1 0 1.30 o. 03
 39.

TABLE IV
RESULTS FOR LIKE LOCATIONS
TRIAL (28/9/71)

LOCATION MOISTURE CONTENT SALT CONTENT

Mean (%) Standard Mean (%) Standard
Deviation Deviation

1 5.63 0.17 1.45 o.os

2 15.65 0.1 5 1.45 0.09
3 1 5. 65 0.14 1.44 o.os
4 15. 63 0.17 1.43 0.09
5 15.65 0.13 1.44 0.1 0
6 15.66 0.17 1 .44 o.os
7 15.63 0.1 6 1. 43 0.07
8 15.66 0.1 5 1 .44 0.07
9 15.62 0.16 1 .44 0.09

OVERALL RESULTS

Moisture Mean (%) 15.64

Standard Deviation 0.15

Salt Mean(%) 1.44

Standard Deviation 0.08
 400

The higher standard deviation for salt and moisture content

compared with the ' slice ' values is a r eflection of change i n the compos-
ition of each location over the time sequence. This again indicates
that the compositional variation over time is more significant than
varia tion across the ribbon at any one instant.

A second experimental run was conducted; 11 consecutive samples

were taken at 1 minute intervals. The results were treated in the same
manner as those in the first t ria l and similar patterns emerged.

The ' slice ' means showed considerable vari ation indicating a
compositional variation with time. The mean values for moisture and
salt content for 1 l ike locations' were close to the grand over a ll mean
values, as in the previous trial. But whereas in the previous run the
variation with time was much greater than that across the s lice as shown
by comparison of the s tandard deviations, this was not borne out in the
second run.

It appears that the second run was of mor e regular composition with
respect to time , but slightly l ess homogeneous across the slice at any
one instant .

Table VI (page 41) gives r esults of noisture and salt content for
the eleven ' slice ' samples taken at specified times.

Table VII (page 42) gives results of moisture and salt content for
'~ locations' for the slice samples taken during this trial.
 41.
TABLE VI
RESULTS FOR INDIVIDUAL SLICES
TRIAL (28/1 0/71)

TIME MOISTURE CONTENT SALT CONTENT

Standard Standard
(minutes) Mean (%) I
Mean (%)
Deviation Deviation

0 15.53 0.05 1. 50 0.07

1 5.63 0.07 1. 52 0.08
2 1 5. 61 0.03 1052 0.07
3 15.58 0.04 1. 55 0.02
4 15.58 0.06 1 .44 0.13
5 15.56 0.09 1. 53 0.06
6 15.61 0.1 0 1. 51 o.08
7 1 5.62 0.04 1. 50 0.05
8 1 5.63 0.06 1 .42 0.1 5
9 15. 52 0.03 1.38 0.06
10 15.58 0.06 1.50 0.11
 42.
TABLE VII
RESULTS FOR LIKE LOCATIONS
TRI.AL (28/1 0/71 )

LOCATION MOISTURE CONTENT SALT CONTENT

Standard Standard
Mean (%) '
Deviation
Mean (%) Deviation

15.59 0.06 1.47 0.08

2 1 5. 61 0.08 1.49 0.09
3 15.59 o.os 1.48 0.1 0
4 15. 58 0.06 1.46 0.08
5 15. 58 o.os 1. 51 0.10
6 15. 57 0.05 1.48 0 . 14
7 15. 58 0.06 1.48 0.09
8 15. 59 0.06 1.52 0.07
9 15.59 0. 09 1.49 0. 12

OVERALL RESULTq_

Moisture Mean (%) 15.59

Standard Deviation 0.07

Salt Mean (%) 1.49

Standard Deviation 0. 1 0
 43.

Analysis of Variance
An analysis of variance was conducted on the two experimental runs.
The factors under observation were :
1) Position - Variations in moisture and salt content influenced by
the location or cross sectional position at which the
sample was taken.
2) Time - Variations in moisture and salt content influenced by
the time a t which samples were taken.

From these results it was found that for variation in moisture

content, a significant portion of the variation could be attributed to a
first order factor interaction of Position-Time. This indicates that
variation is due to a combination of both Position and Time and may not
be fully explained by consideration of each factor separately.

For salt content variation a zero order factor was found to be

significant, this was Time. No first order interaction was present.

It is likely that for moisture variation a zero order factor is also

significant, although this may not be readily examined if a higher order
(first order) interaction has been shown to be statistically significant.

From the previous discussion of Tables III - VII, it may be seen

that variation with time is important. The variance analysis also showed
a high variance value for Time.

Thus it may be concluded that compositional variation is influenced

significantly by Time, and also by the interaction of both Position and
Time.

Because of the difficulty in analysis of the interacting factors

of Position and Time, it was decided that variation in composition should
initially be examined with respect to the base factor - Time.

Note 4
A summary of the Analysis of Variance is presented in Appendix II.
 44.

EXPERIMENT III

Continuous Block Sample

The previous result s showed a product which was varying considerably

with time. By taking a complete section from the extruded ribbon a
continuous sample was obtained. It was then possible to determine the
composition of this section of ribbon at various positions. By determin-
ing the speed of ribbon travel, the distance travelled could be related to
elapsed time. From these results it was possible to determine the rapidity
with which the product composition could change. Variation in composition
within this sample is not caused by intervention of the operator, as the
time over which the sample was taken was a period of less than 30 seconds.
This is in contrast with the previous work, which was sampling over a
long time period, during which operator changes may have been made .

Fig. 6 (page 45) shows a schematic diagram of the continuous block

with the respective moisture content for each 'slice ' expressed
as a percentage. The elapsed time is indica ted on the lower scale.

Samples 1 - 4 were 'slice' samples removed in the normal manner.

Duplicate samples were taken at nine locations as before. 'Slice' samples
were also removed from the mid-point and end of the continuous block.

Samples 5 - 13 were taken at one inch intervals - in this tre atment

the le ading face was scraped in order to remove a representative sample.
Duplicate analyses were conducted. This sampling method differs from
previous techniques. The use of core sampling provided results represent-
ative of the whole 'slice' thickness, whereas the latter is a sample
representative only of the particular f~ce exposed.

From the earlier discussion of errors and analysis of differences

between duplicate results, it was seen that variations in excess of 0.03 %
moisture (the mean of differences between duplicates) could indicate a
real change in composition. Certainly variations greater than the
maximum difference of 0.08 %moisture could be considered significant.

From these results it is seen that compositional changes may occur

rapidly. Over a period of only 5 seconds a change in moisture content
 45

Fig. 5. Locational Sampling Scheme used in Slice Analysis

xx 1
xx xx
2 3
4 inches xx xx xx
4 5 6
xx 7
xx
8
)()(
9

8 inches

Fig. 6. Continuous Block Sample Results

Moisture (%)
15 67 15 63

0 1 2 3 4 ' 5 6 7 8 9 10 11 12

Time Scale (seconds)

Ribbon Speed = 2 inches/second

 46.

of 0.05 %has occurred. Moreover, changes of up to 0.02 %moisture can

occur almost instantaneously , but these are of doubtful significance.

The 'slice' c omposition shows considerable variation. The moisture

cont ent may vary up to 0.25% and the salt content 0.35%. Such a large
variation may not be attributed to experimental sources and must reflect
non-homogeneity of the butter. Compositional changes in a single slice
sample are virtua lly instantaneous, and thus may not be caused by time
dependent variation of process conditions .

Variation in composition a cross the slice may be considered a

function of the mixing capab ility of the machine. Under ideal conditions
complete mixing should produce a butter of homogeneous composition with
respect to a cros s section. Because these conditions can not be met,
variation does occur across a slice.

However, the variation in machine input factors will also give rise
to composit ional variation. If it is considered ideal mixing occurs and
slice composition is homogeneous, then variation in these machine inputs
will result in compositional vari at ion with r espe ct to time.

Analysis of small sections of the ribbon will determine the value of

these short time or ' micro' variations but changing of machine variables
may occur over relatively longer time periods. The effect of one change
may take some time to r each its full value, or it may r each it instantly.
For these changes composition uill vary - either in a gradual change er
a rapid step change res pe ctively.

Thus it is desirable to investigate the 'macro' changes occurring

due to changing conditions, as well as the 'micro' variation. For this
type of investigation the sampling intervals were selected to observe
trends - or changes over an extended time period. A time of one minute
was selected as the interval between samples.

The non-homogeneity of a slice is aless crvariation in comparison

with long term trend changes. Because of this, a single sample was
removed from the ribbon. It was considered that the composition of this
sample, assessed from duplicate analyses, provided a suitable estimate
of the cross-sectional or slice composition at that sampling point.
 47.

The generally low standard deviation across a sli~e supports this

assumption and it has been shown that over several slices the mean
composition, at any of the nine locations examined, is close to the
overall composition of all samples. Thus by sampling at a single point
(plug sample) variations in homogeneity may give rise to a small error
but it is not a consist ent error which could nec essitate compensation.
 48.

EXPERIMENT rl

Compositional Changes Occurring in Butter over an Extended Time Period

Since it has been shown that product variation with time is more
significant than variation within the product 1 (or across the slice) at
any one time 1 this trial sought to investigate changes in composition
with respect to time.

A single plug sample was removed from the ribbon - and this was
considered a representative sample of the product at that specific time.
The plug wa s removed using a l arge 'trier', length of plug= 2.5 11 and
diameter = 1 .5 11 • The plug was placed in n screw top plastic cylindrical
container. By using a moisture impervious cont a iner samples were able
to be stored for some time before analysis.

The sample plug was taken from the middle of the ribbon at the
upper surface. Sixty samples were taken at one minute intervals, and a
duplica te analysis was carried out on each plug sample .

Analysis of Results

The duplicate results for moisture and salt content were averaged
and then treated in the following ways :
(1) Mean a nd standard deviation values were calculated
(2) Graphs of Moisture Content (%) and Salt Content (%) vs. Time were
constructed
(3) Cumula tive sums were determined.* The base value or reference
va lue used was the overall mean value for the trial . Graphs were
plotted for Moisture and Salt Cumulative Sums.
(4) Graphs of Moisture Content (%)vs. Salt Content·(%) were drawn in
some cases. These Scatter Diagrams are useful in detecting the
type of relationship occurring between two variables.

*Note 5
Cumulative Sum Techniques, and Correlation (including scatter
diagrams) are discussed briefly in Appendix III.
 49.

Trial A (24/11/71)
Sixty samples were taken a t intervals of one minute. The samples
were rulnlysed to determine moisture and salt content.

Figs. 7a , 7b (page 50) show the results presented in graphs of

Moisture and Salt Content vs. Time o

Figs . Sa, 8b (page 51) show the Moisture and Salt Cusum Charts.

Discussion of Results

(1) The graphs of Moisture Content and Salt Content vs. Time show
strong relationship between the pattern of moisture and salt variation.
It may be seen that moisture and salt values both rise and fall in a
similar manner, and the magnitude of the change s are comparable.

The correlation coefficient wa s determined between the two sets of

data - Moisture and Salt.
= 60
Number of samples, n
Moisture-Salt Correlation Coefficient 'r' = 0.54
Significance level = 0.001
It is seen that the relationship between moisture and salt cont ent
is highly significant.

(2) The cusum charts however, provide some additional information. The
salt cusum has an initial region of negative slope, indicating the salt
content was below the overall trial mean value. The salt content however,
increases slowly and in the central section of the trial is close to the
overall mean as the graph shows a steady cusum value and has zero gradient.
The final section shows a relatively large change in the mean value as the
graph tends to rise. The changes in slope at times of 12 and 36 minutes
are prominent and may be significant.

The moisture cusum however, does not show such clear trends. Changes
tend to occur more rapidly but are not sustained. Thus although the
moisture and salt values correlate well there appears to be an independent
gradual increase in the salt content over the sampling period.

At the time of this run, other machine factors had not been
 50

Fig. 7a. Graph of Moisture Content (%) vs Time (minutes) (24/11n1 I

Mo~ture (%)
16.0

15.5

15.o ...~~~.....~~--~~~.....~~~....~~....~~~....~~~.._~~--'--~~--~~~--~~~"'-~~--'
0 10 20 30 40 50 60
Time (minutes)

Fig. 7b. Graph of Salt Content (%) vs Time (minutes) (24/ 11/71)
Salt (%)
2.0

1.0 ...~~......~~~.i....~~.....~~...i.~~~.i....~~.....~~....a~----
0 10 20 30
....
40
--~...i.~~~"--~~..i...~~....1
50 60
Time (minutes)

'
 "' ]"c: !?: ]
0
u;
"c:
] ]
"'
E "'
E
i= i=

0 0
"' "'

0
M

t:
0 0
N N
;:, ~
~
;:,
~ ~
.c
u
E "'
.c
a u
u" E
"'~ ~
a 0
t; u"
·;:;
:;; ~
.;
...
"' ·;:;~
.ci
"' 'en"
'" ~
u: E
'"
u: ~
E
E
a ... 0 a ...0 0

u" 0
0 ...0 0 "'! " "'0
u ....
0
0

I I I I
 52.

considered. Later work t ends to indicate that this trend change could
be the result of variation in the salt injection system or possibly some
trend change in &1other machine varia ble.

It is cons idered that the moistUJ:·e content is being influenced by

the injection of the salt slurry. The slurry, being in this case 60 parts
s a l t/40 parts ;rater 1 carries similar quantities of salt and water into the
product. Thus any variation in the dosing volume is reflected as a 'like '
variation in salt content and by asso ciation the variation is refl e cted in
the moisture content.

Salt inj ection is not the sole source of variation in the complex
mechanism of but·cer manufacture, but to gauge the effect of it, if the
1
correlation coefficient r 1
i s squared then the result indicates the
relative influence that the salt slQrry inj ec t i on is having on the
variation in moisture content of the butter.
2
For ' r ' = 0~54 r = 0. 29 or 29%
Thus i n this case 29% of the variation in moisture content is caused by
variat ions in the salt slurry inj e ction system "

Several trials investigating this relationGhip have found a wide

r ange of correl ation coefficients between moisture and salt cont ent . It
is evident that the dependence or relationship between moisture - salt
variations and slurry injection is quite varied. Results show extremely
strong relationship for some sections of a run and yet very little
relat ionshir. at others 'l'ab l e VIII ( page 53) shows the results obtained
from a number of tri als , taken a t different times during the season,

A wide range of va lue s for 'r' is seen. This implies a changing

relationship between moisture and salt values and this is probably due to
the influence of external fuc tor ~ .

The correlation between moisture and salt i s shmm to be significant

in ten of the thirteen trials conducted. In two of the non-significant
trials sample sizes were lower than usual. It has been shown that during
a trial some sections can show little or no relationship between moisture
and salt content wheras the remainder may be highly related.

Two further trials are discussed with r eference to compositional

I

behaviour patterns.
 53.

TABLE VIII
Sffi.1NARY OF RESULTS FOR TH8 CORRELATION COEFFICIENT
BETHEEN BUTT~R HOISTUliJ"S CONT:;;NT AND SALT CONTENT

_::JSample (!)e
Time Period
Considered r
Si gnif-
icance
Noisture
Mean Standard Mean
Salt
Standard
(minutes) Level (%) Deviation (%) Deviation

24/11 /71 60 0- 59 0.54 o. 001 1 5. 48 0 . 09 1. 56 0 .09.

7 /12/71 60 0-5 9 0.70 o. 001 15.63 0 .16 1. 54 0.1 4
21 / 12 / 71 40 0-39 0.56 0 . 001 15 . 41 0 .1 6 1. 52 o. 1 0
7/2/72 40 0- 39 0 .58 0. 001 15.69 0 .1 5 1. 56 0. 09
7/2/72 22 14-35 0. 86 o. 001 1 5. 61 o . 14 1.49 0.1 4
3/3/72 40 0-39 0.35 0. 05 1 5 51
0 0 . 07 1. 51 0. 06
11 /3/72 40 0-39 0. 27 o. 1 1 5 51
0 o. 14 1 .28 0.09
11 /3/72 27 12-38 0 .56 o. 01 1 5 51
0 0.13 1.28 0.09
28/3/72 15 0-1 4 0.29 not sig- 1 5.42 0.10 1. 56 0.11
nificant
6/4/72 40 0-39 0 . 22 not sig- 1 5.41 0.1 9 1.64 0.1 0
nificant
6/4/72 18 0-1 7 0 .64 0. 001 l.5-!E 0 .07 1. 66 o. 11
6/4/72 22 18-·39 0.78 0. 001 15· 26 o . 11 1. 63 0.09
3/5/72 25 0-24 0 .28 not sig- 1 5.2 9 0 . 07 1. 40 0 .06
nificant
16/5/72 40 0-39 0 . 49 o. 001 15.38 o. 12 1. 59 0.06
1 6/5/72 37 2- 38 0.69 0. 001 15.39 o. 12 1. 59 0.06
22/5/72 40 0- 39 o. 31 0. 05 15.39 0.11 1.35 o.os
26/5/72 40 0-39 o.08 not si g- 15.36 0.10 1.19 0.06
nificant
 54.

Trial B (11/3/72)
Forty samples were taken at intervals of one minute and the samples
were analysed in the usual manner.

Figs . 9a, 9b (page 55) show the results presented in graphs of

Moisture and Salt Content vs. Time.

Figs. 10a, 10b (page 56) show the Moisture and Salt Cusum Charts.

Discussion of Results

(1) Comparison of the moisture and salt graphs for Trial B (Figs. 9a,
9b) show the variable relationship which can occur between moisture and
salt content.

The portion of the graph from 12 minutes to 38 minut es was extremely

well correlated ( 1 r 1 = 0.56, significant at 0.001 l evel). The values from
5 minutes to 11 minut es show a vary consist ent s a lt l eve l i n the butter
whereas variation continues to occur in the moisture content, and so the
relationship between moisture and salt is poor"

It is not possible to estab lish a definite reason for this low salt
content. Non-homogeneity of the s a lt slurry could be a contributing
f a ctor. If a l esser amount of salt wa s present in the slurry then the
dosing volume would carry a smaller quantity of salt into the butter and
the relative quantity of water incorporated into the product would
increase.

Changes in salt slurry composition may be caused by

(i) Low level in the tank
(ii) A new charge of slurry may be of different composition.

A further example of low salt content is seen in Trial H (26/5/72).

(See Table VIII (page 53)). The salt content of 1.19% is lower than
usual, and the standard deviation of 0.06 indicates a low variation in
salt content. For this trial the correlation coefficient is very low
('r' = o.os). Thus it is possible that the salt content has a limiting
level below which the injection doses have little effect on the variation
in moisture content, but above this minimum level the relationship is
present.
 5~

Fig. 9a. Graph of Moisture Content (%) vs Time (minutes) (11 /3/72)

Moisture (%)
16.0

15.5

15.oLo------..L..------1~0------.i....-----2~0------...;..------3~0------.._..........._4~0

Time (minutes)

Fig. 9b. Graph of Salt Content (%) vs Time (minutes) (11/3/72)

Salt (%)
2.0

1.5

1.0 Lo---""-----1""'"0---""'----2~0~--""---....i30---..._--~40

Time (minutes)
 56

Fig. 10a. Moisture Cusum Chart (11/3n2l

Cusum (% moisture)
0.4

-0.4

-0.8

- · 0._____________....
10_______________
20_______________
30,_,,____......_______
40
12

Time (minutes)

Fig. 10b. Salt Cusum Chart (11/3/72)

Cusum (% salt)
0.8

-0.4._____....______________..______..______.._____________________~

0 10 20 30 40
Time (minutes)
 57.

The quantity of slurry injected at any one time could be greater

than that which can be effectively handled by the subsequent working and
mixing process. Use of more than one injection section along the working
section or earlier injection could contribute to a less variable product.

(2) Cusum charts for moisture and salt content reveal similar inform-
ation to that discussed on the previous page. The s ections of graph
12 minutes to 38 minutes are virtually identical in their pattern. This
again indicates strong correlation between moisture and salt content,
with minimal ext ernal disturbances upsetting this relationship.

Trial C (6/4/72)
Forty samples were t aken at one minute intervals and analysed in
the normal manner.

Figs. 11a, 11b (page 58) show the results presented in graphs of
Moisture and Salt Content vs. Time.

Fig. 12 (page 59) shows a gra ph of Moisture Content vs. Salt

Content for the trial.

Figs. 13a, 13b (page 60) show the Moisture and Salt Cusum Charts.

Discussion of Results

(1) The results of this trial are of particular interest because of the
marked change in moisture values which has occurred at the time of 18
minutes. The moisture-salt correlation if calculated for the whole run
is not significant, but on separation of the two portions of the graph,
the correlation for each section is highly significant.

This is an important example of a 'step change' in the processing

conditions. In this trial, the step change resulted from changes in
process control settings. An increase in the cream flow rate was made
and at the same time the primary churn speed was increased. The increase
in churn speed more than compensated for the increase in flow rate and
resulted in a rise in the moisture content of the product. This was
independent of the salt content.
 :JO

Fig. 11a. Graph of Moisture Content (%) vs Time (minutes) (6/4/72)

Moisture (%)
16.0

15.oL..-----'-------..&...------'------....,-----.1~----..i...----.....ii....-----
0 10 20 30 40
Time (minutes

Fig. 11 b. Graph of Salt Content (%) vs Time (minutes) (6/4/72)

Salt (%)
2.0

1.00L.----....1~-----1~0------"'-----....l20~----~------3•0------..i...----....i40
Time (minutes)
 59

Fig. 12. Graph of Moisture Content {%) vs Salt Content {%) (6/4/72 )
Moisture (%) (Scatter Diagram)
16.0

15.5

•
15.oL......L-..i..-.i........L--'--...................._ .....__.
1.0 1.5 2.0
Salt (%)
 60

Fig. 13a. Moisture Cusum Chart (6/4/72)

Cusum (% moisture)

-4.0 L..--..L..--....L---.i.---..&...--....i.---.....
0 10 20 30
- -.....---40
Time (minutes)

Fig. 13b. Salt Cusum Chart (6/4/72)

Cusum (% salt)

1.0

-o.2L..___..___L-____.....____-""---------------------------:40
0 10 20 30
T ime (minutes)
 61.

(2) A graph of Moisture Content vs. Salt Content was drawn. Two
distinct portions exist and two separate regression lines may be drawn
through the plotted points. This distinction reflects the two moisture
levels .

(3) The cusum chart of Trial C also shows this very distinct change in
mean moisture value a t 18 minutes. In this moisture chart it is clearly
seen that a significant change in slope has occurred a t 18 minutes.
Subsequent to this change the chart tends to flatten out after 28 minutes
indicating a slight decrease in the process mean.

The salt cusum chart shows a more variable pattern, having several
short term changes in mean value.

However, it is important to see that although the main factor

affecting moisture content at this time is an overriding effect, the
lesser variation in salt content is still r eflected in the moisture
cusum chart.

At 13 - 14 minutes, the salt level drops quite markedly , at this

point also a slight change in slope is seen in the mo isture cusum, a
symp~thetic change.

Similarly , t he moisture and salt cusum charts show a change in mean

at 18 minutes, but the magnitude of the moisture change pr ecludes salt
injection from being the sole causative factor. However, the salt mean
changes again a t 28 minutes and this is reflected as a slight change in
slope also in the moisture cusum.
 62.

The Comparative Magnitude of Moisture and Salt Variation

The relative magnitude of changes in the moisture content and salt

content is important in examination of the relationship between these
two variables.

The salt slurry used during the experi mental runs was composed of
60 part salt/40 parts water by weight. Thus asswning clos e ass ociat ion
of the salt and water in the slurry a known r a tio of salt and water is
incorporat ed into the product. A variation in salt content should also
be seen as a proportionate variation in moisture content if other factors
do not interfere .

Likewi se the standard deviation would be of similar order for both

salt and moi sture content as their r anges of variation would be comparable.

Inspe ction of r esults for Tri a l D (7/12/71) supports these f a cts.

(See Table VIII (page 53)).

Figs. 14a , 14b (page 63 ) show the r esults presented in gra phs of
Mo isture and Salt Content vs. Time.

Fig. 15 (page 64 ) shows a graph of Moisture Content vs. Salt

Content for this tria l .

Figs. 16a, 16b ( page 65) show the Moisture and Sa lt Cusum Charts.

Proportiona l varia tion of Mo isture and Salt values does occur, and
the standard deviation of these factors is a ls o similar indicat i ng
va riations of approximately equal magni tude.

Moisture Mean 15.63 % Salt Mean 1.54 %

Standa rd Deviation 0.16 Standard Deviation 0.14
Correla tion Coefficient 'r' = 0.70 n = 60

If, for example, there was complete interdependence between

moisture and s alt content the correlation coefficient would approach a
value of 1.0 and the proportionate varia tion would a pply in all changes
of salt content and its dependent variable, moisture content.
M
CD g ~ g §
"c: "c:
] i
..
E
..
E
i= i=
 64

Fig. 15. Graph of Moisture Content (%) vs Salt Content (%) (7/12/71)

Moisture (%) (Scatter Diagram)

16.0

• ••
•

15.5

•
15.0 0
1. 1.5 2.0
Salt (%)
 t:
-
N

c 0
N
t:
~
i
(.J
c
E ~
a
:i
.<:
(.J
(.J
E
f
:i a
:i
:;;
·c; ~ (.J

::!:
~
.! ~ ;8
£!
·c;
.~ ,;,
u.. E
u: ~
~
E ~
a
:i
0
,..;
0 N
0
<D
0
0
E
a:i
<D
d
...0 0 m
~
(.J
I I
(.J
'j
 66.

The influence of overriding external factors weakens this relation-

ship9 the correlation coefficient is seen to lessen and expected
proportionality is not presento

The graph of Moisture vs. Salt content also indicated a strong

r e lationshipo If a line is drawn through the cluster of points such that
it approximates to the general trend it is seen that the slope of this
line approaches 38° (a gra dient of a.so). The r egression line calcula t ed
for this data was de scribed by the equat ion
0.79 s + 14.41 where M = butter mo isture content (%)
S =butter s alt content (%)
0.79 - gradient of the line
intercept on the y axis
For an ideal situation of total dependence of moisture on salt the
gradient of this line should correspond to the r atio of water : salt in
the slurry. For the slurry in use - having 60 parts salt : 40 parts
water, by weight, the gradient should be (40/60) = (2/3). This would
r esult in a Moisture - Salt r e l at i onship such that
M .£ s +. c where M butter moisture content (%)
3 S =butter salt content (%)
C - a constant
In the practical situation the preciseness of this r el at ionship does not
exist, the interdependence is more v aried, and the gr adient of the
scatter diagram deviates from the 2;§ predictedo However, over the
consider able number of samples se l ect ed, the comparat ive magnitude of
variation , as seen in both the Scatter Diagram and the graphs of Moisture
and Salt vs. Time; does in fact agree clos ely with theoretical predictions.

The cusurn charts for moisture and salt show marked similarity.
Trends and varia tions are in phase and in the same direction and do
indeed have comparable magnitude.

Other trials runs may not show such strong relationship and this is
likely to result from influences of other factors. With such external
influences minimised, it is probable that the type of behaviour shown in
these results would exist at all times.

It appears at this stage that reduction in the variability of the

product can be achieved by improving the salt injection system.
 67.

EXPERIMENT V
An examination of salt injection pump performance.

Tests were conducted on a salt pump of similar design taken from

the smaller MB 5 model. Results from the tests showed that :
The dosing volune of the pUI:lp varied with pressure required to
pump the fluid. The pressure may be considered as two components
(a) delivery pressure - nt which the slurry is injected into
the churn,
(b) suction pressure - influenced by the level of slurry in the
holding tank.
As the delivery was increased , or the level i n the tank fell, the dosing
volume dropped markedl y. Now, as the machine operator is able to exert
c ontrol on the slurry level in the tank, the main factor to be considered
is the pressure within the working section - that which the pump must act
against to deliver the dosing pulse. Thus the variation in the working
section pressure will affect the dosing and so wil l a lso aff ect moisture
and salt content of the product.

Note 6 Nuoerical results for this test are presented in Appendix I V.

An i mportant corollary of the pump performance is that the Contimab

machine incorporates a water dosing unit for fine control of moisture.
Being the same type of diaphragm pump it is likely tha t this feature also
gives rise t o similar variations i n its output under varying operating
conditions. This could then accentuate or reinforce variation produced
at the salt inj ection stage.

The fact ors which will affect the delivery pr essure ar e det ermined
by those within the working section:
(i) volume of product in the worker - which is a product flow characteristic
(ii) action of the screws within t he worker, and the type of flow which
the butter exhibits in the r egion of the slurry ingress.
It is possible that fluctuations in f l ow will result in fluctuations in
pressure within the worker. An increased flow of cream will result in
increased product in the worker. If this produces a greater pressure
in the worker, then the slurry i njected will be of a lower volume. This
l ower volume of slurry is then distributed throughout the increased we ight
of butter and must r esult in a lower moisture and salt content . It is a
r einforced effect, a higher weight of butter r eceiving a lower dose of slurry.
 68.

The Frequency Distribution of Moisture and Salt Values

From approximately 200 samples analysed a frequency distribution

was calculated for moisture and salt values. The distribution of these
va lues showed similarity to tho 'Normal Distribution ' . Because of the
capability of the oper ator to a lter the process mean values of the butter
composition, this is not a true normal distribution as a de terministic
element is present. It is like ly, however, that should a machine be run
for some time without variat ion of the control factors, the compositional
values would approximate to a normal distribution.

Table IX (page 69) gives the frequency distribution of the 194

results, which were taken from four trials.

Fig. 17 (page 70) gives a graphical presenta tion of these results

in a histogram.

Calculation shovred that the va lues did approximate to a normal

distribution.

Note 7 Details of the calculat ion ar e pre sented in Appendix v.

 69.

TABLE IX
Data for the frequency distribution of Moisture and Salt values
taken from four trial so (25/11 /71 ~ 7 /12/71 , 21 /12/71 , 7 /2/72)

Moisture Class Frequency Salt Class ; Frequency

!

1 5 000 -- 1 5 005 0 1 000 - 1 • 0 5 0

1 5. 05 - 1 5. 1 0 0 1 • 05 - 1 1 0
0 0
15.10 - 15.15 0 1.10-1.15 0
15c15 - 15020 2 1.15-1.20 0
1 5.20 - 150 25 5 1020 - 1.25
15.25 - 15.30 5 1 .25 - 1 030
15.30 - 15.35 8 1 030 - 1 035 3
1 5.35 - 15.40 15 1 035 - 1 .40 12
1 5.40 1 5.45 19 1 . 40 - 1.45 16
1 5.45 15.50 22 1.45 1. 50 33
15.50 - 15.55 29 1050 - 1. 55 37
15 .55 - 15.60 12 1. 55 - 1 . 60 28
15.60 - 15. 6 5 16 1060 - 1.65 26
15.65 15.70 12 1 • 65 - I. 70 24
15.70 - 15.75 22 1.70 - 1.75 5
15°75 - 15.80 11 1.75 - 1 .so 6
1 5. 80 -· 1 5. 8 5 5 1.80 - 1 .85
1 5 085 - 15.90 9 1 085 - 1 • 90 0
1 5 090 - 1 5 • 95 0 1 090 - 1 0 95
1 5 . 95 - 1 6 • 00 0 1 • 95 - 2. 00 0
1 6 . 00 - 1 6 . 05 2 2. 00 - 2. 05 0

194 194

Grand Mean Grand Mean = 1o57

Standard Deviation = 0.17 Standard Deviation= 0.11
N 194 N = 194
 70
Fig. 17a. Frequency Distribution Histogram for Moisture Values
Frequency
40

35

30

25

20

15

10

a....____....._.__.._....,__..__.__.._...,.__.._...,.__.._...,.__...._...,.__..._....i........
--~~

15.15 .20 .25 .30 .35 .40 .45 .50 .55 .60 .65 .70 .75 .80 .85 .90 .95 16.016.05
Moisture Class (%)

Fig. 17b. Frequency Distribution Histogram for Salt Values

Frequency
40

35

30

25

20

15

10

oL-----..c::::.:::1.~--L....JL....L.-~.J....L....J....J.~;;;;;i~~....._~
1.20 .25 .30 .35 .40 .45 .50 .55 .60 .65 .70 .75 .80 .85 .90 .95, 2.0
Salt Class (%)
Conclusions

Thus it is known that in the MC 30 machine:

(1) Variation in product composition is significant over time a lthough

the product at any one instant is of relatively even composition.

(2) The salt pump injection system has been shown to cause up to 74 %
of the variation in mo isture content. The r emainder is a result of
other factors j

(3) Installation of a positive displa cement pump, one having a constant

delivered volume which is independent of delivery pressure could cut down
a portion of this variat ion. However, changes in volume flow will still
influence the moisture and salt content - a s the constant volume of slurry
will be incorpora ted into a varying quantity of product.

The r esults also i mply that present testing frequencies do not

provide a good indication of proc ess changes , as l a rge variations in
moisture and salt content can occur r apidly. The use of duplica t e samples
would reduc e error a.~d the application of statistical techniques such as
control charts would a ssist in be tt er process c·ontrol.
SECTION IV AN INVESTIGATION OF SELECTED PROCESS VARI.ABLES

A definite r e lationship has been established between moisture and

salt contento Variations in performance of the salt slurry injection
system are considered to influence the moisture contento However, only a
portion of the compositional variation can be at tributed to this source ,
and so furth er investigatory work was initiated with the aim of assessing
other possible sources of process variation ,

The factors investigated were those considered to be a possible

major source of variation in product composition " Owing to the complex
nature of continuous butter manufacture a great many base factors and
interacting factors of higher order could be sources of compositional
variation. Because of the restricted time available for this investigation
only a limit ed number of factors could be considered.

By comparing the patterns of variation of the selected variables,

with those of the butter composition parameters it will be possible to
determine if relationships do occur. From this information the type and
strength of such a relationship may be determined .

A. ComRosi t iona l Factors

Rm·r materi a l or process factors which could influence butter

moisture or salt content were studied. The possible r ange of values of
the selected factors needed to be such that a real variation could be
caus ed in the butter composition if the factor value was not s teady.

The factors investigated were

( 1 ) Fat content of the cream
(2) Temperature of the cream
(3) Titratable acidity of the cream
(4) Product flow r ate
(5) Pressure within the working section

These factors were investigated and then patterns of variation were

compared with the characteristic moisture and salt variation o If a
particular factor was seen to behave in a consistent manner with little
variation, while over the same time period the moisture-salt variation
was still present at its typical level, this factor was not considered
 73°

to be a prime source of product variation.

Cream Characteristics

Cream characteristics play an important role in continuous churning.

The effects of variation in cream temperature, fat content and titratable
acidity have been discussed in Section I and their quantitative influence
has also been included.

In examining cream factors, it is important to determine if any

sustained change in their value has occurredo Minor fluctuations of
short duration are likely to have little effect on butter composition
because of the system's 'capacitance' which will tend to exert a 'damping'
effect. A change of duration e.go greater than 30 seconds may begin to
have an important effect on churning, A sustained step change in the
level of a cream factor will cause a change in the process and this will
be reflect ed in the product composition. Similarly, cycling of these
variables will result in cycling of the product composition factor3, if the
cycle has sufficient amplitude and time period. The causes of variation
in cream characteristics may be rather complex, some basic causes will
be discussed brieflyo

Cream Temperature
Variations in cream temperature may be initiated at a number of
points in the process. Defective or faulty operations may themselves
initiate temperature variations or sustain those variations already
present.

Faulty cooling after cream vacreation may give rise to temperature

change. However, these variations are not likely to be sustained unless
subsequent operations during cream holding are faulty,

Temperature variation may be caused during the holding period in

the vat prior to churning. Inoperative, or inadequate agitation may
allow stratification of the cream. The subsequent cooling operation
to compensate for the latent heat of crystallisation may not be evenly
applied. In addition, regions near the cooling surface may drop in
temperature more rapidly and this cream will increase in viscosity
resulting in reduced heat transfer to the bulk of the cream. If this
 74.

occurs uneven cooling is also likely to result . Not only may the final
crernn temperature be variable if uneven cooling occurs , but the nature
of the f a t may vary due to differing treatment applied to that fat . This
is also likely to contribut e t o variable churning properties,

Cream Fat Content

A major caus e of variation in cream fat cont ent is the stratification
which results from faulty agitation. In addition regions close to the
cooling surface may a lso have compositional differences if compar ed with
the bulk cream . Adequ&te cream agitation should minimise this source of
variation.

Titratable Acidity
Variations in titratable acidity of the cream may be the result of
defective operation of the neutralising unit if automatic control is used.
Batch neutralisation can give rise to variation in a cidity of the cream ,
but in both cases adequate agitation should preclude variation.
 75.

Trial D ( 7/ 2/72 )
Temperature of the Cream

Procedure
The cream temperature was measure d at the balance tank prior to
the mono-pump o A Copper-Const antan thermocouple was ins erted through
a s ection of stainless steel tubing which was placed down the side of the
balance tank into the centre of the pipe opening which led to the cream
pump.

An ice junction was used as the refe r ence tempera ture and the
output voltage was recorded on a Varian Continuous Chart Recorder, Model
G-1 5-1 •
Scale range - 1 mv
Chart speed - 1 "/2 . 5 minutes

The cream temperatur e was recorded over a period of 5 hours 25 min-

utes, commencing at startup of production and running during processing of
the firs t vat and a lso during the first 30 minutes of use of the second
vato During a selected portion of the temperature measurement run 40
samples of cream and butter were taken at one mi nut e interva ls "

Results
Fi g. 18 (page 76 ) shows t he graph of Scale Response of TeIJperature
vs. Time for cream flowing through the bal ance tank during the period in
which cream and butter samples were taken.

It has been shown in the Ca libration of the Varian Recorder tha t

the Response (Sca l e reading) i s directly proportional to t he temperature.
Thus the graph (Fig. 18) is in effect a graph of Temperature vs. Time.

From the time-temperature relationship it is seen that during the

period in which butter and cream samples were taken the cream temperature
was constant.
Scale reading = 4s.5 + 0.5 units
Temperature = 54°F + 0.3°F
Thus variation occurring in the product parameters of moisture and salt
content may not be attributed to variation in cream temperature for this
run.
N
r::
N
2 0
M

,c
~

l
e
;::
~

·c:
::>
~

~
"'
·~
~
,
~

m
~
....~
~
0
N
~
u
0
~

ii'
t;
cxi
,;,
u:

I'

c:
::>
~

~ I
0
~ ~
0
N
0
M
0
'<I'
0
"'
0
...
0 0
"'
0
~
·.:
"' "'
>
 77
Fig. 19a. Graph of Moisture Content (%) vs Time (minutes) (7/2/72)
Moisture (%)

16.0

1s.0J..._ __._ _ _...__ _...i.---~20,_.--..i...--~30~--,..--~40

0 10
Time (minutes)

Fig. 19b. Graph of Salt Content (%) vs Time (minutes) (7 /2/72)

Salt (%)
2.0

1.oL.--_.,_..;.......,ji....--.-..i..--.....1i---""---....i..---..._--~40
0 10 20 30
Time (minutes)
 Figs 0 19a, 19b (page 77) show the results presented as graphs of
Moisture and Salt Content vso Time. Thes e samples were taken during the
stated period in the cream temperature time trial.

It is seen that the characteristic variation of moisture and salt

content still occurred. A significant correlation 'r' = 0.58 was
calculated for the data.

Moisture Mean 15.69 1a Salt Mean = 1 • 56 71a

Standard Deviation 0.15 Standard Deviation 0.09

Although cream temperature can give rise to variation in product

composition, as shown previously, the characteristic moisture and salt
variation has still occurred even when cream temperature is stable.
Thus it would seem that cream temperature is not one of the major variables
being sought which gives rise to the characteristic process variation.

Description of Crea!'J Temperature Graph

The time-temperature graph during the run shows little fluctuation
in cream temperature o Some variation is seen as a direct result of flow
alterations 9 vat changes and similar oper ato r intervention. These
variations are of short duration and small magnitudeo Thus they will tend
to have only a minimal influence on the process.

Generally cream temperature is seen to be stable and would not

appear to be a major source of process variation o
Trial E (7/2/72)
Fat Content of the Cream

Procedure
The cream was sampled at the balance tank prior to the cream pump
(Fig. 2 (page 17)). This position was chosen because of the relative
ease in obtaining samples. Sampling at any other position could have
required modification to the pipe layout " In add ition , consideration
of the mixing occurring in the balance tank and cream pump showed that
sampling prior to the b&lance tank would not provide cream repr esent at ive
of that flowing to the continuous churn.

The crorun was remo ved at one minute int ervals from the balance
tank and put into screw top bottles. The samples wer e refrigera ted
until analysed. The Babcock cream testing procedure was used and duplicate
fat analyses were carried out on each cream sample , The sample volUI:J.e
ta.ken was 150 mls. By using a r el atively l ar ge volume it is hoped that
errors due to sampling of non-homogeneous substances are mini mised.
The mean value of the duplicate Babcock r esult s was calculated.

It is known that the mo isture content of butter can be altered if

a significant change i n the fat content occurs. To deterCTine if a relation-
ship existed between variations in the fat coLltent of cre am and t he moisture
content of the butter a correlation coefficient was calculated for the two
time series of results. However, because of the capacitance of the system,
a time delay will be exporienced between the times of sampling the crean ,
and the butter which is manuf a ctured from that cream. The time delay may
be considered as the sun of four components:
(i) Time taken between sampling of cream and the exit of that cream
from the balance tank.
(ii) Time ta.ken for the cream to pass to the mono-pump and then into
the primary churn.
(iii) The time e l apsed between entry of the cream to the churn and its
extrusion as the product.
(iv) The time ta.ken for transport of the butter down the chute to the
sampling location.
 so.

Calculation of the Time Delay

Conditions assumed:
Churn Capacity = 9000 lb/hr butter
Cr e Cream = 40 %fat
Butter Composition - 15.8 %Moisutre
1. 5 %Salt
1 .2 'fb Curd
Approximate quantity of butter
in churn (capacity) = 350 lb
Specific Gravity Cream (4o% fat) = 0.995 (31 )
;:::s 1• 0

(i) Residence time in balance tank

Fig. 2 (page 17) shows a diagram of the ta.lance tank and the dimensions
are included .
Diameter of balance tank = 2 f ee t
Height of balance tank = 1.5 fe et
2
Volume of balance tank = rrD h
4
2
rr • 2 • ( 1 • 5) rt 3
4

= 4.7 rt 3
Cream Ca pacity = 295 lb
Now cream 40 %fat
100 lb cream~ 40 lb butter fat
Added Moisture + Salt + Curd = 1s.5 % of the total butter
~ 40 lb butter fat = 81 .5 %of the total butter
Thus butter produced from
1 00 lb cream = 40 x 100 lb
81 . 5

49 lb =
Thus 1 unit of butter requires 2 units of cream by weight
Now butter flow = 9000 lb/hr
= 9000 .ll. x 1 hr
hr 60 min

= 150 lb/minute
Thus cream required to produce
this butter = 150 lb/hr butter X 2 uni ts cream
1 unit butter
= 300 lb/minute
 BL

Thus capacity of balance tank 300 lb

Flow of cream through balance tank = 300 lb/minute
Thus average cream residence time = 1 minute

(ii) Residence time in pipe line between balance tank and churn

Length of pipe 33 feet

Diameter of pipe 2 inches
2
Area of pipe 1l?_ square inches
4

~ ~ f~~2 )
2 2
( rr! ) inches X (
1

= l!l.1)
(144 ft
2

Flow through pipe = 300 lb/min

Cream specific weight = (62.4 lb ). x 0.995
\ ft 3
3
3
( oo ;~n) x ( (-62-.-4-~-o-.9-9-5"T"")i-~- )
= 4.s ft 3/min
Since velocity of flow Flow rate
Areu of flow

The velocity in pipe ~ ( 4. 8 !i: ) X ( b:!:'. ) X

6
( ~:t4 ;t 2)

= 3.7 ft/sec
Length of pipe = 33 ft
Residence time in pipe ( 33 ft) X ( 1 sec )
3.7 ft
9 seconds (ii)

(iii) Residence time in churn

The assumption is made that the butter exhibits plug flow in the
churn.
Primary Churn
It is not possible to calculate the residence time for this section,
as the proportion of the primary churn which is filled is unknown. An
estimate was made from a trial, of the time taken to travel through the
line from the balance tank and through the primary churn.
Estimated time = 12 seconds
 820

With a full line this tim0 may be increa sed ,

Time allowed = 15 seconds (inclusive of (ii))
Working section
It was estimated that the working section contained 150-180 lb of
butter.
Now as butter production 1 50 lb/min
Then the working section residence time is approxi mate ly 1 minute .

(iv) Time taken to travel down the chute to -position of sampling

Lengt h of chute t r nvelled = 9 feet

Speed of butt er = 2 inches/sec
~ Time of travel = 54 seconds

Thus tota l time delay between sampling of cream and sampling of the
related butter : -
Time - Cream balance tank 1 minute
Pipe line and pri~ary churn 0.25 minutes
Working section = minute
Chute travel = minute

~ot a l time = 3.25 minutes

Calculation of the Correlation Coefficient

Consider if no time delay existed between the series of fat values

and the se~ies of moisture values. The correlation coeffici ent would be
calculatec from thes e two time series.

FAT CONTENT VALUES

0 2 3 38 39

MOISTURE CONTENT VALUES

0 2 3 38 39

:f a time delay of 1 minute existed then as the moisture content is

the vuiable which lags, we would then compare the second moisture value
with ~he first fat value, since the butter moisture value would be
dependent on the cream sample taken one minute earlier. Similarly, the
2, 3, 4, ••• 38 minute values for f at content are compared with the
3, L, 5, ••• 39 minute values for moisture.
 83.

0 2 37 38 FAT

MOISTURE
2 3 38 39

The moisture series is in effect displaced by 1 tine Lnit relative to the

fat series. The direction of displacement is such that it compensates
for the time delay occurring in the systema

This displacemen t procedure may be continued to 1J.11Y desired level.

It should be noted that as each displacement is effected, the sample size
for calculation of the correlation coefficient is reduced by onea It is
important that the true sample size be consider ed in determination of
statistical significancea If a r el at ionship does exist between the t wo
variables then a significan t correla tion coeffici ~nt will be found.
A maximum value of the correla tion coefficient should occur at the
displacement level where the two series have been matched corr ect ly -
that is when the system time delay has been fully compensateda If the
time delay is not an integral number of minutes then it is possible that
more than one significant correlation coeffici ent may occur, at either
side of the true value ,

By determining a series of correla tion coefficients for successive

displacement values of the two series it is possible to:

(1) determine the extent of relationship between moisture content

variation and crean fat content variation - if such a relationship
exists

(2) determine the average time delay occurring between the sampling of
the cream and butter. In this way the residence time in the churn
can be found from data obtained during operation of the machine.
Procedure
40 samples of cream and butter were taken at one minute intervals
and these were analysed for the relevant factors , A correlation analysis
was carried out on the butter moisture and cream fat values, a displace-
ment analysis was usedo

Figo 2():1.(page 85 ) shows the graph of Fat Content VSo Time for the
trial o

Table X (page 86 ) shows the r esults from the Corre l ation analysis 0

Discussion of Results
It was shown in the previous discussion that in general, an
increase in cream fat cont ent causes an increase in the moisture content
of butter, and thus a positive correlation coefficient is expe cted if a
relationshi p exists between these two variableso

From the tabulated results (Table X), three values are statistic-
ally significant - at 1 , 9, 16 minutes displacement o The 9 minut e
displacement value is significant to the 0 .01 level of confidence , but
the correlation coefficient is negative and this is contrary to the
expected resulto The coefficients at 1 , 16 minutes are positive, which
conforms with the theoretical result but they are less significant, a
confidence level of 0.10o However, none of these results is in the r egion
of the predicted delay time, although this may be due to incorrect
assumptions in the calcul at ion of the delay time.

Table XI (page 87), gives a summary of significant r esults obtained

during throe experimental runs. The methods of sampling and analysis
were identical for each trialo

Results from another trial (6/4/72), are presented as a whole, then

as two separate sections. (This is sinilar to their treatment when
moisture/salt relationships were being considered and is due to the
distinct moisture levels occurring during the run.)

The results presented in Table XI do not show a definite trend or

relationship between moisture and fat values. The displacement values
 85

Fig. 20a. Graph of Cream Fat Content (%) vs Time (minutes) (7 /2/72)

Fat (%)
42.0

36.0L.----...L------..L..------L...-----2•0------.i...------3~0------------~40
0 10 Time (minutes)

Fig. 20b. Graph of Cream Acidity (% Lactic Acid) vs Time (minutes) (6/4/72)

Acidity (% Lactic Acid)

0.12

0.10

0.08

L-__...____.___....J..__--J.____. ___ ~ ___.___~

0.06 0 10 20 30 40
Time (minutes)
 86.

TABLE X
RESULTS OF THE CORRELATION COE FF IC IENT BETWEEN
VALUES OF CREAN FAT COHTENT (zfl AND BUTrER MOISTUR.E CONTENT (%)
FOR SUCCESSIVE V.ALUES OF SERIES DISPLACEMENT

TriaJ. ( 7 /2172)

Butter Mois-'c1;;.:c-e Content Mean (%) 15.69

Standard Deviation 0.15

Cream Fat Content

Standard Deviation 0.8

DISPLACEMENT COIDl.F.LATION srIPLE LEVEL OF

VALUE COEFFICIENT n r·rn<ICiuTCE

0 -0.149 40
o. 261 39 0.10
2 Oo 11 9 38
3 -0. 072 37
4 -0. 042 36
5 -0.071 35
6 0 . 21 0 34
7 0 . 253 33
8 -0. 1 59 32
9 -0.484 3i o. 01
10 -0.093 30
11 0 .1 32 29
12 0.083 28
13 0.123 27
14 -0.098 26
15 -0.1 89 25
16 0.365 24
17 0.237 23
18 0.069 22
19 -0.201 21
20 0.114 20
 TABLE XI
SUMMARY OF RESULTS FOR THE CORRELATION COEFFICIENT
BETWEEN BUTTER MOISTURE CONTENT AND CREAM FAT CONTENT

SampleiDisplace- ! Signif - i Moisture Fat

Date Size , ment i r l icance ! Mean Standard Mean i Standard
i
!
(n)
I
Value
I
Level (%) \Deviation
;
(%) : Deviation
7/2/12 I
!
40 !
i
1 ! 0.26 0.1 0 15.69 i
I
. 0.1 5 39o5 0.81
I
I
I
I
I
9 -0.48 0.01 .i'
1

I
16 0.36 0.1 0
II
I
i
!
3/3/72 I 40 ! 0 -0.36 0.05 15.50 ! 0.07 39.5 0.56
:
I 1 -0.36 0.05 I
! I
I
j
4 -0.47 0.01
I
i
I

6/4/12 I 40 8 0.29 0.1 0 15.40 II 0.20 39.6 0.44

II I 9 0.31 0.1 0 l
I
I
18 j
I,(o -1 2 0.41 0.10 1 5.22 I 0.07 39.4 0.36
I
7) I I I

TABLE XIII
SUMMARY OF RESULTS FOR THE CORRELATION COEFFICIENT
BETWEEN BUTTER MOISTURE CONTENT AND CREAN ACIDITY
!Sample Displace- Signif- i Moisture Acidity
Date 1
; Size I ment r icance j Mean istandard Mean i Standard
(n) \ Value Level i
(%) ;Devia tionl (%) Deviation I
l
6/4/72 40 29 -0.57 0.05
'
115.40
I
I
;
0.20 0.002 0.003

18 \ 12 0.62 0.1 0 !1 5.21 I 0.07 0.083 0.003

(0-17) I
I0.082
I
I
22 I 7 -0.48 0.05 15. 58 I
I
0.11 0.003
(18-39)1 i I

I, 9 -0.70 0.02 I I
j
12 o.so o. 01
I
I I
Ii 14 I
I 0.54 0.1 0
i
t !
!

(Cream Acidity is expressed as% lactic acid.)

 88.

do not show a consistent pattern for achieving correct compensation for

the time delay between the two series. Some significant values occur
in advance of the calculated time difference. If this calculated value
is reduced to a minimum, still being compatible with fixed transport
time factors but minimising the variable components, some significant
values are still occurring in advance of the calculated time value.

Analysis of the correspondi ng standard deviations for moisture content

and fat content for each run does not provide additional information.
There does not appear to be dependence or relationship between the magni-
tude of variation in the butter moisture content and tha t of the cream
fat content.

Discussion of Errors
.An analysis of errors is importo.nt in assessing the results of
this trial. Determination of f at content by the Bab cock method is
subject to errors, some of which may be quantitatively assessed .
Duplicate analyses were made for each sample. The differences between
the duplicate results provide information regarding the accuracy and
repeatability of the test. From the fat 8.Jlalyses the differences
between duplicate results were nnalysedc
Number of samples, n = 79
Mean difference , µ = o. 5 as percent age f at
Standard deviation of
differences , s = 0.56

Table XII (page 89), shows the frequency distribution of differ-

ences between f at duplica t e results.

It may be seen that the majority of such differences are less

than 1.0 %fat, although five values exceed this level and are as high
as 2.0 %fat. This magnitude of 'difference' is significant in comparison
with the variation occurring in the cream fat content values as determined
from the mean of duplicate samples. The mean difference is large in
comparison with the standard deviation of the f at values. Thus the
variation in cream fat values may not correctly indicate actual variation
in composition due to the considerable errors involved in the estimntion
of the fnt content. Because of this, comparison of the fat values with
those of butter moisture content will not give reliable information
TABLE XII
FREQUENCY DISTRIBUTION FOR DIFFERENCES BETWEEN
DUPLICATE ANALYSES FOR CREAM FAT CONTENT

DIFFERENCE (% FA'I
1
) FREQUENCY
BETVESN DUPLICATES

0 26
0 .1 0
0. 2 5
0. 3 4
0. 4 0
0 .5 25
o. 6 0
0.7 2
o.s 0
0. 9 0
1. 0 12

1. 5 3
2. 0 _2_
-13_

= 79
Sampl e s i ze n
Hean Difference o. 5 7~
Standard Deviation of Differ en ces = 0 . 56

TABLE XIV
FREQUENCY DISTRIBUTION FOR DIFFERENCES BETWEEN
DUPLICATE ANALYSES FOR CREAN ACIDITY

DIFFERENCES (% LACTIC AC ID) FREQUENCY

BETWEEN DUPLICATES

0 22
0.005 11
0.01

Sample size n = 40
Hean Difference = 0.003
Standard Deviation of Differences = 0.004
 900

concerning any int e r-relationshipo

The analytical errors have become significant because of the

rela tively small changes in cream f at content < Experimental methods
of sampling, analysis and e specially re ading of the r esult, do not allow
a sufficiently a ccurate answer to be obtained using t he Babcock t es to
If the variation in fat cont ent was considerab ly gre at er than the
differences beuveen duplicate r esult s then the analytica l errors would
be less important o

Conclusion

(1) It has not been shmvn that va ria tion in cream f at cont ent contrib-
utes to the cha r a ct eristic variat ion in moisture or salt content of
buttero .Analytical errors could cont ribute significantly to the ' a pparent'
varia tion in f a t cont ent and thus an incorrect pattern of fat values
obtained.

(2) Use of mor e a ccura t e testing me t hods could reduce analytica l error
to an a ccept able level. The r el at ionship betw een cream fat cont ent and
moisture content could then be r eass essed.

(3) Sampling errors could become significant if more accurate

analytical techniques a r e used. It would be important to ensure that
bulk samples were homogeneous prior to remova l of sampl es for arlalysi s .
 91.

Trial C - Cream Acidity (6/4/72)

Procedure
The titratable a cidity was determined for 40 cre3.II1 samples taken
at one minute intervals , using the method discussed previously. A
correlation analysis was carries out between the two series of values -
cream acidity and butter moisture. The displacement technique was used.

Fig. 11 a (page 58) shows the r esults present ed as a graph of

Moisture vs. Time for the trial.

Fig. 20b (page 8 5) shows the r e sults presented as a graph of

Titratable Acidity, calculated as l actic acid, vs. Time for the trial.

Table XIII (page 87) presents a summary of significant results

obtained in the correlat ion analysis.

Due to a significant proc ess cha nge the results are presented in
three sections in a similar manner to the moisture results for this
trial.

The summary of results for corr e l ation between moisture content

and titratable a cidity shows certain cha racteristics which may be likened
to those of the moisture-fat content analysis.
(i) The time displacement shows no particular recurring or cyclic
value for significance in correlation .
(ii) Both positive and negative corre l at ion coefficients occur. .An
increase in acidity tends to cause an increase in moisture content and
thus it was expected that a positive correlation would occur between the
moisture and acidity results.
(iii) A maximum correlation coefficient is not seen as evidence of
correct serie s displacement. In a ddition, the significant values do not
occur at the same level of time delay as those significant values in the
moisture-fat analysis. This would have been expected had a correct time
displacement occurred.

Discussion of Errors
The importance of analytical errors in the determination of cream
acidity is similar to that concerning cream fat content analysis.
 92 0

If the ~ a ctua l' va~iation in a cidity i s no t high then the experi-

mental error be comes significant and it can contribute to the 'apparent'
varia tion in t he acidity va lues .

The diff erences be t we en duplica t e r esults wer e a lso ana lysed for
t his tria l. Tab 1.e XIV ( page 89 ) present s t he frequency distribution of
differ ence s be tween duplic a t e a cidity ana lyse s .

Number of s ampl es, n 40

Mean dif fe r ence, µ 0.003 (%l a ctic a cid)
Standar d C:.evint ion of
dif feren ces, s 0.004
This compar es with the a cidity dete r mined a s t he mean of duplica t e analyses.
Nw:Jber of sampl es = 40
Over all mean a cidi ty = 0.082 (%l a ctic a ci d)
Standard deviat i on 0. 003

From t he tabulated resul ts, the diff erences of up to 0. 01 %l a ctic

a cid may contr ibute a major porti on of the variat i on i n the r esults.

deviation of a cidi ty values, and thi s Kas seen a l so in t he cream fat

content r e s ult s .

(1) It ha s not been s hown that varia ti on i n the ti trat ab l e a cidity of

cream contribut es to the char a cteristic va ria ti on i n moisture or s alt
cont ent of butter.

(2) Analytica l errors could contribut e to 'apparent varia tion' in

titra t able acidity va lue and thus an incorrect seri es of the values
obta ined. This is a r esult of the low variation in the titratable acidity
s eri es. The pos sible experimental error is l arge when compared with the
magnitude of variation.

(3) Reassessment of the butter moisture - cream acidity relationship

could be practical if the experimental errors could be reduced. Consid-
eration should be given to obtaining homogeneous samples as variations
in composition could become important if experimental accuracy was improved.
 93.

B. Churn Variables

Two process variables were investigated under normal running

conditions
(1) Linear extrusion speed of the but ter ribbon>
(2) Pressure exerted by the dosing pump in inj e ctin~ the salt slurry
into the working section.

These factors have been previously discussed with reference to

variation in salt injection. Fluctuations in pressure have been shown to
affect the salt slurry pump performance. It was considered that flow
variations could influence pressure conditions within the working section
and thus the pressure that the pump must overcome in order to inject the
slurry dose. (ref. page 67)

1 • Linear Extrusion Speed

The linear extrusion speed of the ribbon is a 'flow' parameter. It
reflects the volume of output, as the cross-sectional area of the ribbon
is constant and is determined by the size of the extrusion head orifice.
An increase in the cream flow rat e should result in a corresponding increase
in extrusion speed which should be directly proportional unless density
changes occur. Major density changes are not likely to occur as the
Contimab machine incorporates a vacuum working section. Howeve r, slight
fluctuations in linear speed may be a reflection of events in the working
section rather than the cream flow r ate.

2. Slurry Injection Pressure

Both cream flow rate which is rel a ted to the quantity of product
in the worker, and flow behaviour in the working section could affect the
flow characteristics in the region of slurry ingress. This is likely to
influence the pressure at this point and so the performance of the slurry
pump. Thus these factors may be sources of variation in pump performance
and thus in the salt and moisture content.

Measurement of these two factors together with the butter moisture

and salt values would enable comparison of behavioural patterns to
determine interrelationships .
Measurement Techniques

Linear Extrusion Speed

A toothed wh~el was rotated by the butter ribbon flowing from the
extrusion head of the churn. The spindle of the wheel was connected
through a gear train to a tachometer. The pulse rate of the tachometer
is proportional to the rate of revolution of the wheel and thus is related
to ~he ribbon speed. The pulse output was summed for an accurate time
interval by an electronic timer-counter and the result displayed digit-
ally. Thus the summed result taken o'V\:lr a constant time interval
indicated the relative ribbon speed. A series of readings wa s taken over
a suitable time period. Because only the speed trends and variations
were under consideration calculation to determine the actual ribbon
speed was not necessary but it may be detetmined in the following manner.

Tooth wheel diameter =

Circumference of wheel = rr • 7 5k inches
24 inches
revolution of tachomet er produces 112 pulses
revolution of toothed whee l produces, through gearing, 25
revolutions of tachometer.
Thus 1 revolution of wheel = 25 X 112 pulses
= 2800 p'\l.lses
Thus 2800 pulses = 24 inches run
1 pulse 2.4.._ inches run
2800

This conversion factor may be used to determine the ribbon speed.

Sample calculation
Speed mean = 257 pulses/second
Ribbon speed 257 l ~ inches/second
2800

= 2.2 inches/second
= 11 feet/minute

Apparatus
Advance Timer-Counter Model SC3
Frequency of speed readings = 26/minute
Summation interval = 1 second
Salt Dosi~1:!8]2_Pulse PreSSl.1£..§.
A diaphragm joint was placed in the slurry injection line. Move-
ment of the diaphragm as a result of applied pressure was transmitted by
a fluid link to a Bourdon tube. A link rod was attached to the moving end
of .thE: Bourdon tube. The link was then connected to an inductive
displacement transducer and the output was derived through use of a
displa cement converter 2.mpl ifier. The output from the amplifier was
displayed on a..1 oscil loscope and a permanent record of each pulse made
by filming the stationary oscilloscope trace.
Boirrdon tube 0 - 1 20 psi
Displacement Transducer Philips PR-9314A/10 (displacement± 10mm)
Amplifier converter - Philips PR-9309 Nomimi.l Output + 1 v.
Oscilloscope - Telequipment S52
Camera -· 35nm with motor driven spools
FreQuency of pulse - 40 pulses/minute
The height of the trace was measured on the film using a Lupe Magnifier
( 5X) fitted 1-ri th a millimeter scale. 'l'he pulse heights were read
manually . Under normal conditions pressures ranged up to 60 psig.

Photographs of the apparatus used are presented in Figs. 21a, 21b

(page 96).

,frocedll.J'.'_Q.
Continuou3 read ings of ribbon extrus ion speed , and slurry dosing
pulse pressu.re were taken over a period of 40 minutes. Samples of butter
were taken at 1 minute intervals over the same period. Results for each
variable uere gr nplrnd
(i) Variable values vs, time
(ii) Cusum chart

Control changes made by the machine operator were recorded. A time

delay will exist ~etween the a ctual time of change and the time at which
the change is seen in the product composition 3 and it should be considered
in interpretation of results.
The data obtained was highly variable and many short time fluctua-
tions were seen. To remove this short duration 'noise' from the graphs
the data was modified and the mean value of eight consecutive values was
determined. This modified value was considered as the data value and the
graphs of Variable Value vs. Time and the Cusum chart were derived from it.
This applies to both ribbon speed and slurry pump pressure results.
 96 .
Fig . 21a Apparatus for Measurement of Ribbon Speed .

Fig . 21 b Apparatus for Measurement of Salt Slurry Injection Pressure .

 97.

Discussion of Results

Trial C (6/4/72)
Results were obt a ined for ribbon speed and butter composition.

Churn oper ating conditions

Vari ab l e ' Cream Pump Salt Pump Worker Churn

Se tting Setting (rpm) Speed
Time (min.) \

0 3.25 25 28 .5
18 3.5 31 increase
28 29 290 5

Mea..11 Standard Devi ation

Moisture ('.;t) 1 5.41 0.10

Salt (%) 1.56 o. 11
Ribbon fmeed 12.0 0.3
(ft min)

Moisture-Salt Correlation
Time 0-17 minutes (n = 18) 'r' 0.64 significant at 0.01 level
Time 18-39 minutes (n = 22) 'r' 0.78 significant at 0.001 level

Figs. 11 a , 11b (page 58) show the results presented as graphs of

Moisture and Salt Content vs. Time.

Figs. 1 3a , 13b (page 60) shm·T the Moisture and Sa lt Cusum charts .

Fig. 22 (page 98) shows a graph of Ribbon Speed vs. Time.

Fig. 23 (page 99) shows the Ribbon Speed Cusurn Chart.

The Moisture and Salt charact eristics of this run have been discussed
earlier and particular reference was made to the significant change in
moisture mean. The ribbon speed cusum chart shows a similar change in
mean at a time of approxiamtely 18 minutes. This is indicated by a
marked change in slope.
 "'"'
~1

~ l
e
i=

~
~~
~

:;
<
>
>
~ 0

"'
-s.
______>
~
-~
_?
~
-=-=--
~

~
\
-------- ..?

~ I
-<;.::.----
-?
__.......::-=- I
-?
l
-~
0
N I
/
'
-c ~\.._ I
..-->-
.;>
~

~
?
-
\
"!_
!!!

¥ -=::::::-~
--
1 -:.<
~
m
E
i=
~

'ii >-
~
~
8
~
~ ~~
a:!!_
"'c - -- :=

~
0
:ll
a:
0
~ ~
~

~
t
-s ~

<
'ii
2
~
~
8
~
J
~

l
"'
-.?--
8... 0
8 0 80
"'"' "' "'
N N

:
 §
ij
\ ~

~
,::

f
~'
\ \
\ \
\,
\
'\

>\
./.
\
z.,.....
~ .....
'
"' \
z
\

'

:;a

'
'·,
l
l
'-

,/'
;
/ '
/

;;;
r:::
I
J
~
"
.l!
u
E
~
/
\
\
~

/
u

I
"' I
I
~ //
...,.; /~

•
...
r
J

,/
, ..,,.

r'~
I
§
E
(
,/
g ..
, /

,
u
·-~ 0
~ ...8I 8
0

~ ~I §
I
l.
MAS~!:. l.11 Y:..R!>ITY
 1 oo.

From the churn operating data it is seen that at this time the
cream pump flow rate was increased by nearly &'fa and thus the volume flow
was increased proportiona tely. Calculation showed th~ t the mean ribbon
speed increased by a similar amount.

Cr eam pump speed (i) 3.25

(ii) 3..!5_
Difference ~
7.7 %based on original v alue

Ribbon speed (i) 259 pulses/unit time

(ii) 280 pulses/unit time
Differ ence _gj_ pulses/unit time
= 8 .1 %based on origina l value

At the same time, the primary churn speed and the salt pump setting
were altered in order to maintain the butter composition. A nett incr ease
in moisture cont ent resulted and a slight nett decr ease in salt content
occurred.

It is thus seen that the step change in cream flow was reflected
accurately by the ribbon speed measurement. The increase in mean moisture
content of o.44)b would probably account for the slightly greater incr ease
in ribbon spe ed.
 10L

Trial F (16/5/72)
Results were obtained for ribbon speed, slurry pump pressure and
butter compositiono

Variable ! Cr eam Punp Salt Pur::tp Worker Churn

Speed
Time (mi~:Jl Setting Setting (rpm)
I
0 )oO 35 29
4 31
12 29
15 32
No further change

Mean Standard Deviation

Noisture (;0) 15 .38 Oo12

Salt (76) 1. 59 0.06
Ribbon speed (ft/min) 11. 0 0.3
Pump pressure (units) 85 5

Moisture-Sa lt Correla tion 'r ' = Oo49 significant at 0.001 level

n 40

Figs. 24a, 24b (page 102) show the results pr esented as graphs of
Jl!ois ture and Salt Content vs. Time.

Figs. 25a, 25b (page 103) show the Moisture and Salt Cusum Charts.

Fig. 26 (page 104) shows the graph of Ribbon Speed vs. Time.

Fig. 27 (page 105) shows the graph of Salt Pump Pressure vs. Time.

Fig. 28 (page 106) shows the Ribbon Speed Cusum Chart .

Fig. 29 (page 107) shows the Salt Pump Pressure Cusum Chart.
 102

Fig. 24a. Graph of Moisture Content (% ) vs Time (minutes) (16 /5n 2l

Moisture (%)
16.0

15.5

15.ooL------..L..------1~0------.!.-----~20~-----'-----~3~0-------'------4~0
Time (minutes)

Fig. 24b. Graph of Salt Content (%) vs Time (minutes) (16/ 5/72)
Sa lt (%)
2.0

10 20 30 40
Time (minutes:
 103

Fig. 25a. Moisture Cusum Chart ( 16/5/72)

Cusum (% moisture)
1.0

-1.o ·L
0
------.i...----....i.------"'------"""------------..-.------------.-
10 20 30 40
Time (minutes)

Fig. 25b. Salt Cusum Chart (16/5/72)

Cusum (% salt)

0.8

- 0 · 4 0~----....1.------~1~0------"------2~0~----.i...----~3~0-------------4~0
Time (minutes)
 0
M

<'<
t::
~

1
1
~
;::

~
8
~ S!
~
].
a:
~
!t
"'c 0

a:""""
0
-a
~
.;
N

'"
u:

'g

~
1
a:
!t
"' 0

~
0 0 0
...
0
~ 0
N
 N

~
-~
.s
~
;::

"§
,~
~
~
j
~

~
Q.
E
rf
Ji
0
-a
~

:::i
u: '"

]
~
~

~
0
~ ..
0
...
0
~ 0
0
c
.8
£
a:

:ilI
 (
J

,.
/
/
/

/
/
 1 080

The graphs of Moisture and Salt vs o Time accurately show the effect
of machine changes o At 4 minutes a change in the worker speed initiates
an increase in moisture content, this is then re-corrected at 12 minutes
and the process returns to steady moisture level at 15 minutes o The
change in salt pump setting at 15 minutes has resulted in a near propor-
tionate change in the butter salt content.

Salt pump setting (i) 35

(ii) .R
Difference -2.
= 8. 5 %based on original value

Salt content (mean) (i) 1 • 65 7;

(ii) .1ill. %
Difference 0.12 %
= 7.2 %based on original value

An a pparent change in the salt content at 5 minutes has not been

caused by an operator change in salt pump setting or cream flow rate.
It is possible that a change in pump performance has occurred. For
example, malfunction of the diaphragm due to fouling of the seat by salt
particles may affect performance. There has not been any sustained
change in either ribbon speed or slurry pump pressure at this stage.

The cusum charts for moisture and salt again show good relationship
and they reflect process changes ac curate ly.

The graph of ribbon speed data showed s ome cyclic forms, and this
is al so seen on the cusum chart. The cycle duration varies from 8 - 11
minutes and is characterised by a rapid rise of nearly 10 %in the level
of the variable followed by a slow decrease to near the starting value.
This cycle commences at approximately 4 minutes, the time at which an
alteration was made in worker speed. It is possible that the initial
increase in worker speed causes an increase in ribbon speed, but which
slowly resumes its original level which is probably dependent on the
butter flow rate. However, a decrease in worker speed at 12 minutes
does not have the converse effect and decrea se the ribbon speed as might
be expected. Later results showed this cyclic form to occur independ-
ently of operator control changes in worker speed. It is possible that
this cycle is caused by flow characteristics.
 1·09.

The graph of pressure data shows some recurring regions of low

pressure where approximately ten cons ecut ive pulses are a t a markedly
lower pressure than the rema inder. These aberra tions appear to be
related t o the cycling of the ribbon speed as the times of occurrence are
in close agreement. It is possible t hat this is caused when the packer
screws commence operation rui.d draw the butter down the chute. The incr ease
in speed could result in lowering of the back pressure in the worker.
Although the speed change is relativ el y sustained the pressure rapidly
resumes its normal level. Perhaps this is because the flow pattern in
the worker is rapidly restored to normal despite the continuing pull of
the screws in the ribbon . However, the cycle time of up to 8 minutes is
far in excess of the time taken to fill the forming chamber and extrude
one carton of butt er. Thus although this phenomenon may be rela ted to
the packing unit the source is not the regular pn.cking cycle.

Other causes may be r e l a ted to the flow input or cream pump

performance and perhaps to flow pn.tterns in the worker .

Pressure level (i) 90 units

(ii) fil.. uni t s
Difference --3. uni ts
= 10 %change based on original value

Thus under the conditions pr esent, tho r elationship between salt

content , pump setting and injection pressure is we ll defined.
 11 Oo

Trial G (22/5/72)
Results were obtained for ribbon speed, slurry pump pressure and
butter composition.

Churn operating conditions

Variable Cream Pump Salt Pump Worker Churn

Setting Setting (rpm) Speed
Time (min.)

0 4.2 4.1 38
2 32
No further change

Mean Standard Deviation

Moisture (%) 15.39 0.10

Salt (%) 1 .1 9 0.06
Ribbon speed (ft/min) 13. 9 1.0
Pump pressure (units) 79 5

Moisture-Salt Correlation 'r' = 0.31 significant at 0.05 level

n = 40

Figs. 30a , 30b (page 111) show the results presented as graphs of
Moisture and Salt Content vs. Time.

Figs. 31a , 31b (page 112) show the Moisture and Salt Cusum Chartso

Fig. 32 (page 113) shows the graph of Ribbon Speed vso Time.

Fig. 33 (page 114) shows the graph of Salt Pump Pressure vso Time.

Fig. 34 (page 115) shows the Ribbon Speed Cusum Chart.

Figo 35 (page 116) shows the Salt Pump Pressure Cusum Charto
 111
Fig. 30a. Graph of Moisture Content (%) vs Time (minutes) (22/5/72)

Moisture (%)
16.0

15.5

15.0 0
10 20 30 40
Time (minutes)

Fig. 30b. Graph of Salt Content (%) vs Time (minutes) (22/5/72)

Salt (%)

2.0

______....,______2_0______..._______ ____....
40
1 0
·

0
1-----~------~

10
3~0------._

Time (minutes)
 112

Fig. 31a. Moisture Cusum Chart (22/5/72)

Cusum (% moisture)

0.4

-0.8

-1.oL.---'----L---iL....--..L..--....L---&.....--..__ __.
0 10 20 30 40
Time (minutes)

Fig. 31b. Salt Cusum Chart (22/5/72)

Cusum (% salt)
0.6

-o.4L---'---..J..--...1.--....J~--._--~---'----;:4o
0 10 20 30
Time (minutes)
 ~-------

- --_,.
::-:-__._,_
~--=-

---- --~--...
-- /

-=-- --- ~

--
......--~
.-.c. =:c.::.~ ~

---- ~
:!
0-

'} ~~
i
~
;::

\
~ /

(
·----..._> 0

< "'

~
/ \
_,,/'

{
?
)
~
l
;>
'-,,
,,
,,,.-- /

">·\ ~
\
.->
{

/)
< ,,
;:;;
t:: (
~ t'

.,
c
z
.( '

~""
i
~
;::

~
.
-~

5 '> ''· ~

~ ,.,,,.
"'
~" ' ··
c.
E
~ ....:.
,f '.I
~
0 '?
-a ,)
~ !
,.; '!
M
,,
>
,,
'"
u:
('
(
~
~
<--:
<
~ ~
/
0
0 0
~
0
~ ~ 0
"' "'
l
"'

l
~
~
~
ui - - - -g
--~ g
0
0
"/ -- I
E
~
u
 117.

During the trial no operator changes were made to the churn after
2 minutes , thus variations occurring are the result of process variation
inherent in the system.

The moisture and salt graphs again show relationship in their

variation; their correlation is significant.

At 2 minutes a major change in worker speed was made, this is seen

also as a change in ribbon speed but is not sustained. It would appear
that the ribbon speed is only temporarily affected by changes in worker
speed. The prominent and sustained changes in ribbon speed are likely to
result from flow rate changes.

The moisture and salt cusum charts show that a process change has
occurred at 20 minutes. The moisture and salt cusum values begin to
increase. At this time a reduction in the pressure l evel has occurred
and it is probable that an increase in salt dosage volume resulted
because of this. Such an increase in the dosage volume will cause the
changes in moisture and salt content .

Pressure level (i) 81. 0 units

(ii) Xi:..Q units
Difference _&..:..Q units
= 7 • 5 fa change based on original value

Salt Content (i) 1 .35 %

(ii) .h.22. %
Difference Q:L fa
= 15 % change based on original value

This change was not sustained for longer than 4 minutes but the
available data does not provide an explanation for this.

The moisture and salt values show further change at 23 and 26

minutes although significant changes do not occur in either ribbon speed
or pump pressure.

The ribbon speed shows the occurrence of a short time cycle -

which exhibits quite large differences in ribbon speed. This is due to
 11 80

the behaviour of the ribbon during this run. The ribbon was forming
ripples down its length and these were initiated at the extrusion head.
A high production rate was the cause of this and the increased flow
resist&.""lce was giving rise to temporary build up or slowing of the flow
until the resistance was exceeded. This short time effect was not shown
in the slurry pump pressure measurement as might have been expected.

The ribbon speed cycle previously discussed, one of 8 - 11 minutes

duration is again pres ent and is superimposed upon the short time cycle.
The cycle in this instance has not been initiated by operator changes in
the worker speed . 'rhis cycle may be caused by :
(i) variation in the working speed which is dependent on the system and
as yet , not controllable.
(ii) butter flow rat e variations.
The magnitude of the speed change is not large and is approximately
5-1 ry/o.

The pressure results do not show the same regions of low pressure
as in Trial F (16/5/72) which coincided with the ribbon speed cycling.
Generally the pressure level is steady and shows only minor variation.

The latter stage of the speed cusum chart corresponded to a similar

pattern in the butter moisture cusum chart. A positive correlation of
this type would suggest worker speed as the causa tive factor. This
factor exhibits a positive correlation with moisture content whereas
cream flow rate shows a negative relationship. (This has been discussed
in Section I (page 23)) .
 11 9.

Trial H (26/5/72)
Results were obtained for butter ribbon speed 9 slurry pump pressure
and butter composition.

Churn oper ating conditio11.§_

Variable Crear:i Pump Salt Pump Worker Churn

Time (min.) Setting Setting (rpm) Spe ed

0 4.2 37 44
No further change

Mean Standard Deviation

Moisture (%) 1 5.36 0.1 0

Salt ( ~&) 1.1 9 0.06
Ribbon speed (ft/min) 14.5 0. 8
Pump pressure (units) 92 7

Moisture-Salt Corre l a tion 'r' = 0.08 not significant

n = 40

Figs. 36a 9 36b (page 120) show the r esults presented as graphs
of Moisture and Salt Content vs. Time .

Figs. 37a 7 37b ( page 121 ) show the I.foisturo and Salt CusUir. Charts.

Fig. 38 (page 122) shows the graph of Ribbon Speed vs. Time.

Fig . 39 (page 123) shows the graph of Salt Pump Pressure vs. Time.

Fig. 40 (page 124) shows t he Ribbon Speed Cusum Chart.

Fig. 41 (page 125) shows the Salt Pump Pressure Cusum Chart.
 120

Fig. 36a. Graph of Moisture Content (%) vs Time (minutes) (26/5n2l

Moisture (%)
16·0

15.0L.--•--..,.jl---..1---....L~--.i..---:3~0--------:.40
0 10 20
Time (minutes)

Fig. 36b. Graph of Salt Content (%) vs Time (minutes) (26/5n2)

Salt (%)
2.0

1.5

1.oL.._ _..__ _~1~0--_.---!20~--.i....--~30~--......--:40

0 Time (minutes)
 121

Fig. 37a. Moisture Cusum Chart (26/5/72)

Cusum (% moisture)
1.2

0.8

0.4

- o. 2 •0--------------1·0--------------2·0--------------3·0---------------40
Time (minutes)

Fig. 37b . Salt Cusum Chart (26/5/72)

Cusum (% salt)
0.6

0.4

- 0.4

10 20 30 40
Time (minutes)
 -----------
______/
~

-- __-:s;::
::::>-
-
~o: __

~--:.;;;::------=-=-

<> J

----~

__ ?,

-=:::::....---- ----
2
 g

0
N

"'
 7

~
u
-·
I

E
a
8
l
"'c
~a:

0
[ii
I
J!!
u

,
c.
E
rf
 126.

The cusum chart for salt content and pressure indicate the expected
relationship between these two variables. The salt cusum shows a general
decrease followed by an increase which continues after 20 minutes. The
pressure cusum shows the converse effect as expected. The changes seen
in these variables are slow and the cause may be .related to a slow change
in pump performance. Flow va riations seen in ribbon speed do not
correspond with moisture and salt cusums.

Over the whole run the average pressure drops by approximately

1S'fo , the mean salt content incre ases by a comparable amount, although
a proportionate relationship was not really expected.
Salt Content (i) 1 .15 %
(ii) ~ %
Difference ~
= 17 % change based on original value

Pressure level (i) 99 units

(ii) §1. units
Difference .112, units
= 18 %change based on original value

The moisture cusum chart appears to be related to the ribbon speed

cusum and the correlation i s positive. This tends to indicate, as in the
previous discussion of Trial G (22/5/72) the variation is occurring as a
result of changing worker spe e d rather than cream flow rat e . Evidence
is also seen of the cyclic form in ribbon speed which again is superimposed
upon variation caus ed by other factors,

These r esults show variation in slurry injection to have a minor

effect on variation in moisture content. The salt content is generally
low and has low standard deviation. This allows a dominant effect to be
exerted by additional factors , in this instance variations in worker
speed could be responsible.
 127.

Discussion

Thus three different situations have been seen to occur.

Trial C ( 6/4f72) and Trial F ( 16/5/72)

Operator changes in machine variables have provided a dominant
influence tending to determine the salt and moisture va lues . Ribbon
speed is well related to butter and cream flow r ates.

Trial G (22/5/72)
Outside influences are minimal and varia tion occurs from within
the system. The expected pressure-salt relationship is seen to occur
and influences both salt and moisture cont ent. A positive relationship
is evident between ribbon speed and butter moisture content in latt er
stages of the trial and suggests variation in worker speed as the
causative factor . This relationship however, is not particularly strong
and has only a minor influence on moisture cont ent.

Trial H (26 / 5/72)

The rela tionship between ribbon speed and moisture content is more
definite and appears to be a major factor controlling the moisture content
for this trial. The variation in worker speed is suggested as the
caus at ive factor. The moisture cont ent is virtually independent of the
salt content. Pressure variation is the dominant influence on varying
salt content.

From these results it would appear that varying emphasis occurs

on different r el at ionships .

(i) Moisture and salt variation may be highly related when the effect
of other factors is minimal.

(ii) A strong pressure-salt relationship may exist which also influences

moisture content. The variation in pressure introduces an external
factor.

(iii) Moisture-salt interdependence may be low. The moisture content

is primarily related to ribbon speed effects, while the salt content is
related to pressure variation.
 128.

It is seen that the ovorriding factors are those machine variables

controlled by the operator. If this influence is absent internal pres-
sure and ribbon speed effects become evident. Under conditions of
minimal variation, the salt-moisture relationship may become the prime
source of compositional variation. These conditions have actually been
present during several runs and thus to aim for minimising of the
external factors is surely not an impractical task.
 1290

S:::irne similarities were observed between graphs of Speed and Pressureo

A correlation analysis was carried out and a displac ement series used.
Displacenent of speed relative to pressure and pressure r elative to
speed was made.

The in terpretat ion of these results did not reveal a consistent

relationship between spoed and pressure. It is probable that speed and
pressure are ind eed. related, but influi:mce of external conditions
overrides these interacting effects. Removal or minimising of these
external effects may cause a spee~-pressure r e lationship to have real
significance when running under l ess variable conditions .

The correlation analysis s:1owod both positive and negative

coefficients , a~though these were confined to separ ate :i:'uns .

A negative correlation may occur if a constant quantity of butter

is passing through the worker, a speed,._:!.nc_r ea_§~ wi ll tend to _red~ the
llit~rn.a)_pr es sur.:El as the effective cross sectional area of flow will tend
to be reduced.

A positive corrc l at io ~1 may occur if a var ying quan ti ty of butter

is pass~_n g thr ou.::;:1 the 1·rorking sectim. Ju1 _in_C]'.'_?_n_~e. i n butter flow uill
.5-n cr._2R?2. t he ribbon speed, and a lso tend to r esult i n a pressure j.~~·

This assumes that butt ~ v density is not changing significantly.

Thus, the chu:::n conditions are a complex set of f a ctors. The

dominance of any one facto::- is the crucial condition and before an
accurate model may be constructed a considerab le quantity of further
information is essential .
 130.

Conclusions

It has been shown that relationship does exist between the chlU'n
characteristics of ribbon speed, salt pump injection pressure and the
butt er compositional factors of moistlU'e and salt content.

However, such relationships are complex taking several forms,

each of which may be determined by the extent of influence of external
factors. The data available was able to describe the type of relation-
ship existing between the f a ctors but there was insuffici ent information
to determine the reason for such a relationship; why it should occur in
preference to other forms.

To fully understand the mechanism which determines butter compos-

ition additional factors must be studied. Only by knowledge of these
influences can an a ccurate model be built to attempt explanation of
continuous butter manufacture. Such a model is of great assistance
if a control system is to be devised for the process of continuous
buttermaking.
 1 31 •
SECTION V SUGGESTIONS FOR FURTHER STUDY

Cream flow rate and primary churn beater speed are prime factors
in influencing butter composition. Accurate measurement of these
factors is needed to continue investigation of the continuous butter-
making system.

The speed of the screws in the working section has also been shown
to exert an important influence on butter composition and must be
considered in further study of this topic.

The difficulty in accurate measurement of these factors has

deterred investigation up till the present time. The problems asso ciated
with this may be resolved by use of suitable analogue-digital techniques.
The vast quantity of information available may be recorded in such a
form that presentation of the data could be made directly to computing
facilities without the time consuming transcription by human operators.

Accuracy of Measurement reguired

Since it is desirable to detect changes of less than one standard

deviation in moisture and salt content, these values should be measured
to an accuracy of at least 0.05%, expressed as actual moisture or salt
content.

This level of accuracy may be related back to the quantitative

effects of variation in churn factors (previously discussed in Section
I (page 15 ff)). It has been mentioned that a variation in primary
churn speed of 100 r.p.m. may result in a butter moisture content change
of up to 0.5%. Thus the requirement for measurement of primary churn
speed approaches an a ccuracy of ± 10 r.p.m. ( assuming approximate
proportionality). This level of accuracy is quite demanding as the
base churn speed may reach 2000 r.p.m.

Cream flow rate also has an important influence on butter moisture

content and under normal running conditions a change in flow rate of
1o% may give rise to a butter moisture content change of up to 0.3%.
This suggests 'that measurement of cream flow rate must be sufficiently
accurate to detect changes of 1-2% in volume flow. Under normal
conditions the cream flow rate is approximately 300 pounds per minute
 132.

(page 8 0) and thus the accuracy required in measurement of cream flow

approaches ± 3 pounds per minute o

The speed of the screws in the working section i s a major factor

used in process control. Alteration of butter moisture content of up
to 0.1% may be achieved by a change of only 1-2 r.p.m. in the worker
speed. The accuracy required in measurement of worker speed would
thus be in the r egion of± 0.5 r.p.m. as change s of 0.05% butter moisture
content may be considered significant.

TABLE XV ESTIMATED ACCURACY REQUIRED FOR MEASUREMENT

OF CHURN VARIABLES
Factor Primary Churn Cream Flow Worker Screw
(Beater) Rate Speed

Base Value 1 000 - 1800 rpm 300 lb/min 30 - 60 rpm

Accuracy
Required + 10 rpm ± 3 lb/mi n + 0.5 rpm

: • % Change
Detected + 0.5 % + 1 % ± 1 %
 133.
SECTION VI
Conclusions

This study has investigated variation in composition of butter

manufactured by the Contimab MC 30 continuous churn during commercial
production. From the initial study it was shown that the behaviour of
the salt injection system contributed markedly to the patterns of
variation found in product composition.

Further process factors, suggested by the results from the

initial study , were examined with respect to their influence on product
composition. Relationships were found between a number of factors
including butter ribbon speed, pressure within the working section,
flow rate and the variation in butter composition. The type of relatio-
ship varied considerably - depending upon the presence of external
factors. The nature of the dominant external factors is not yet fully
known.

Knowledge of the buttermaking process is not sufficiently complete

to enable building of a model to describe the total system. Further
work may allow building of a successful model which would allow improved
process control through application of a suitable control system. If
this is realised, then the r eduction in process variation achieved from
control could prove greatly beneficial to the economic operation of
continuous butter manufacture.
 134.
APPENDIX I

1) Weight changes during cooling of empty moisture dishes, dried

overnight in 100°c. oven.

Dish Weight change during period (grams)

( o. 5 - 1. 0 hr) ( 1 • 0 - 1 • 5 hr) ( 1 • 5 - 2 0 hr)
0

+0.0002 +0.0002 -0. 0002

2 +0.0002 +0.0002 -0. 0001
3 -0.0006 +o. 0001 0.0000
4 +o. 0001 +o.0004 -0. 0001
5 +0.0006 -0· 0001 -0.0003
6 +0.0003 +0.0002 -0.0002
7 +0.0007 0.0000 -0.0002
8 +0.0006 0.0000 -0.0004

Standard -0.0005 -0.0005 -0. 0001

Weight change was considered to be at an acceptable level after

1.5 hrs. cooling.

2) Weight changes during cooling of dish + butter sample after

moisture removal ,

Dish Weight change during period (grams)

( 1 • 0 - 1 • 5 hr) ( 1 • 5 - 2. 0 hr) (2. 0 - 2. 5 hr)

+0.0007 0. 0000 0.0000

2 +0.0005 +o. 0002 -0. 0001
3 +0.0018 +0.0001 0.0000
4 +0.0005 +O. 0001 +o. 0001
5 +0.0008 +0.0006 +0.0004
6 +0.0004 +0.0003 0.0000

Standard +0.0005 +0.0007 0.0000

Weight change was considered to be at an acceptable level after

2.0 - 2.5 hrs. cooling.
 135.
APPENDIX II
SUMMARY OF ANALYSIS OF V At1IANCE

Factors ConsJd££._()_d.
( 1) P·-Posi ti on. Nine positions Here analysed in duplicate. The
level of this factor = 9" Degrees of freedorJ = 8.
(2) T-Time - the number of samples taken "
For run (a) -· 14 slice samples
(t) - 11 slice samples
The f ac tor level (a) 14
(b) - 11
Degrees of freedom(a) - 13
(b) - 1 0
(3) R-Repetitions the duplicate r esults give rise to some variation
and this factor was annlysed separ ate l y. Howeve1., the mean sum of
sqw:~res of this factor uns low in comparison with others. Indeed it
should not pro7ide a signi ficant source of variation. The higher order
factors inc orporating this source f actor , a lso had loll' mean sum of
sq_unres i n comparison with others. The zero and higher order f a ctors
which included R uere summed as the Resj_duoJ_ for the purposes of this
9l1a lysi s.

NOISTURE COlTT~Wr

FACT OR I SSQ i DF I I13Q

..-.,,_.. _., __-- . . . ~- -=-· - ~- -~ -.:_j .---~---
L .'· 2
... - -
p 0.04.996 8 o. 00624 2.96
T 5.13669 13 0.39513 187 .27
Px'l' Oo 56400 1 04 o. 00542 2.57
RESIDUAL Oo 26545 126 Q ,, 00211
TOTAL 6.01610 251
---- -·-

v1 = 101). v2 = 126

fO. 01 = 1.66 s 2 = 2.57

1
(60, 120 )
s 2
2
Thus PxT is significant at level f0.01.
 136.

Trial (28/9/.7£)

SALT CONTENT
2
FACTOR SSQ DF MSQ ~

p 0.01489 8 o. 00186 1.69

T 1 .26494 13 0.00973 8.85
PxT o. 31761 104 0.00305 2.77
RESIDUAL 0.14008 126 o. 0011
TOTAL 1 • 7352 251

PxT Interaction
v1 = 104 v2 = 126
fO. 01 = 1 .66 s 2 = 2.77
1
(60 ,120)
s2
2
Thus PxT is s ignificnnt at l evel f0.01.

Trial (28/10/71)

MOISTURE CONTENT
2
FACTOR SSQ DF MSQ ~

p 0.02684 8 0.00333 1.164

T 0.25378 10 0.02538 8.874
PxT 0.45636 80 0.00570 1.99
RESIDUAL 0.23090 99 0.00286
TOTAL 0.96788 197

PxT Interaction
v1 = 80 v2 = 99
f0.01 = 1 .84 s 2 = 1. 99
(60,60) .J_
s 2
2
Thus PxT is significant at level f0.01.
 137.

Trial (28/1 0/71)

SALT CONTENT
FACTOR SSQ DF MSQ ~2

p 0.06496 8 0.00812
T 0.491 01 10 o. 0491 0 1.457
PxT Oe66856 80 0.00835
RESIDUAL 0. 67611 99 0.00683
TOTAL 1. 90064 197

PxT Interaction
v1 = 80 v2 = 99
ro.05 = 1.4 s 2 = 1 .457
1
(75 , 1 00)
s 2
2
Thus PxT is significant at level f0.05.

Comment

The significance of PxT interaction indicates that a portion of

the variation in results may be explained only by a combination of
both position and time variables. The importance of time is clear
because of the extremely high mean of squares values assigned to it,
although its statistical significance may not be validly tested as a
higher order interaction (PxT) has proved significanto
 13s.
APPENDIX III
.ANALYSIS OF RESULTS USING CUMULATIVE SUM (CUSUM) TECIDUQUES (32)

The basic procedure for calculation of Cumulative Sums consists

of subtracting a constant quantity (eogo a target value ) from each va lue
in the series and summing the resulting differences. The successive
accumulated differences termed the Cumulative Sums may be graphed to
give a 'Cusum Chart' .

The cusum technique has the a dvantage of detecting changes in the

mean value of the process variable. Historical trends are clearly
shown and the onset of any change in the process variable is evident.
A reliable estimate of the current average value may be made rapidly.
The cusum technique contrasts with quality cont~ol methods such as the
1
Shewhart' chart where actual values are plotted. This type of chart
is used to detect process values which fall outside warning and/or
tolerance values. It may not readily show changes in the process mean
value when fluctuations occur .

The main application of cusum techniques has been in industrial

quality control. Process or product variab les may be determined at
regular intervals and compared with pre-specified levels. Movements
away from the target values may be d etected r eadily and their magnitudes
determined so as to enable corrective action.

Determination of Cumulative Sums and Chart Plotting

Consider a set of r esults x , x , x , ••• xn obtained at equal

1 2 3
time intervals, and a reference value (which may be zero) K.

If the calculation s1 = (x 1-K) is performed, a difference is

obtained , and the origin of measurement is in fact changed to K.
K may be a target value to which the variable (x) is supposed to
approach.

If consecutive differences are calculated and then summed

s2 = (x1-K) + (x2-K)
s3 = (x1-K) + (x2-K) + (x 3-K) and so on.
then sn = sn-1 + (xn-K)
= x1 + x2 + ~ + xn - nK
 139.

Successive values of s1 ,
S , s ,
2 3
0 0 0 S are plotted at equal time
n
intervals to construct the cusum chart .
(1) If the values of x are greater than K, (the target or r eference
value), a graph of positive slope will be drawn.
(2) Similarly , a graph of negative slope will be drmm if x is less
than K. The value or degree of the slope indicates the magnitude
of the difference between x and K.
(3) If x approaches closely to K, then a steady line of slope 0 will
be drawn.

Thus any prolonged change in the mean value of x will result in

a change in the slope of the plotted cusum. By determining the relative
changes in slope significance of the change in mean may be determined.
Suitable methods have been devised for testing of significance. An
example is the V-mask test.

Cusum charts have been used in this project to indicate changes

in the mean value of the variable. Generally the r eference value (K)
has been s e lected as the grand mean of the set of r esults, hence the
initial and final plotted values clos ely approach zero. The cusum
chart has been constructed and regions of changing slope have been
investigated to determine if important changes in the mean value have
occurred.

In some cases comparison has been made between cusum charts of

different variab les to determine if a relationship exists between these
variables. The cusum chart is well suited to this, as changes in slope
may be readily examined.

It is important in the use of cusum charts to select suitable

graph scales. If quantitative significance tests are being conducted
the scales must be correctly sized. If such t ests are not being used,
then the scale sizing is leas critical, the criterion being the
sensitivity and ability to show an important change in the mean value
of the process variable.
 140.

Correlation

Correlation analysis is a technique used to investigate the type

and extent of relationship which exists between two variables. If
paired values of two variables x and y are plotted as a point (x,y)
the plotted points form a scatter diagram. Fig. 42 (page 141) shows
the three possibilities -
(i) marked relationship
(ii) a degree of interdependence
(iii) independence
reflected in a scatter diagram.

A correlation coefficient may be calculated as a quantitative

guide to the degree of dependence between x and y. The correlation
coefficient does not indicate which variable is dependent and which is
independent. Such distinction is a problem of interpretation and may
require additional information.

Statistical methods for determination of correlation are helpful

techniques to evaluate relationships but interpretation of results is
an important factor. Interdependency may be indicated by correlation
analysis yet in fact this may be refuted by subsequent interpretation
and investigation. Correlation analysis has been used in many sections
of the project. Interpretation of such results is found in the relevant
discussion for each section.
 141
Fig. 42. Scatter Diagram - Graph of Y vs X

• ••
••
•••
••• • ••
••• •
•
••
• •••••
• ••

(a) marked relationship

x

y
• • •
• • ••
•
• •• • •
• • • •
• • •
• • • •
• • •
•••

(b) a degree of interdependence

x
y

• • •
•• • • • • •
• • •
• • • •
•
• • •
• ••
• • •

x
(c) independence
 142.

§.£!.1.T sr,~Y INJECTION FUNP PERFORMAlifCE

The punp frori o. ContiTiab HB 5 was tGsted for performance under

varying conditions. Tho pump uas obtained froo the Dairy Research
Institute.

Procedure
The dosing quantity for 10 cons ecutive pulse s was determined for
a nuro.ber of cycle :. : " The fluid used uas \nter o A val ve was placed on
the deli very line and. r.·ras varied to alter the applied deli very head.
A pressure gauge Y C.s also plac ed in tho line to determine the applied
delivery heado The pressure moo.surGd. uas that pressure in ex cess of
the static head of tho po.rticular systelilo A variable diaLJ.eter pulley
was fi ttod to 0nab le nlteration of the punping r ate (pulses/second).

A sori oo of trinls uas conducted and pump performance determined

1LYJ.der varying c onditions of fluid level in the r eservoir, deli very head
and pUBping :'at ·J.

A sunrr1a ry of those r e sults is presented below. In all cases ,

tho qu11..YJ.t i ty for each s o1Jple has been collected fro 11 1 0 consecutive
pulses o.nd has boGn ex:p:;:ess od [lS we i ght in gramso

Mean I
Fluid So.mp l e Ql' r·l b ty I
Standard
G""'
~ ..l --

S11oed olivery Level i n Size ( sun of 10 Deviation

i oO
.cessuTe Re ser voir pulses)
~---·~ ( ~siJ ___
(1) llor mnl j 0
---- ---~_.... __,,,...

High
----- ---
20
-- (grao.s)

35 °5 0.63

---- -~--· ·

(2) Nor mal 10 High 20 3201 1.3

(3) High 0 Low 20 36°4 1.3

·-·
(4) High 10 Low 10 26.9 1.1
 143.

Significance tests were carried out to compare performances under

different conditions, using the criterion

t = (x1-x2)

sPJ ~1 + _L
n2

2 2
where s (pooled variances) = (n -1 )s + (n -1 )s
p 1 1 2 2
111 + n2 - 2

where t -· n value of the t distribution

x1 ' x2 - ncnn value of sawple 1 ' 2
n1 , n2 size of SO.l!lple 1 , 2
s1 , s2 - snnpl c variances ( 1 ' 2)
v = (n1+n 2 - 2) degree s of fre edom.

Sa@ple Calculation
To test if the differ ence in pressure between Trial (1) and (2)
has significantly chang~d thu dosine quentity.

v = n1 + n2 - 2
= 38

to . 01 -- 2 . 576

s = (111 - 1)s12 + (n2-1 )s22

p
n1 + n 2 - 2

2 2
= 19(0. 63) + 19(1 . 3)

(20 + 20 - 2)

= 19(0.4 + 1 . 6)

38

sp =
 144.

t = (x1 -x2)

sP,~
1
+
I
n2

0.316

t = 1 0.8

Thus there is a significant difference in the mean dosage given

by the pump, between trials (1) and (2). A decrease in dosing quantity
occurs if the delivery head is increased. This was shown to be true
for both low and high levels in the reservoir.

It was shown that a decrease in reservoir level decreased the

dosage markedly when the delivery head was at 10 psi. It should be
noted that the decrease occurred even though the rate of pumping was
increased" Compare trials (2) and (4). This was not true when the
delivery head was 0 psi. The increased rate of pumping has increased
the dosing quantity in spite of lower reservoir level. This is
probably a result of improved sealing of the diaphragm at higher
pumping rate. It is likely that the dosing qua.Dtity would have
decreased as the reservoir level dropped, had the pump speed remained
the same and it should be noted that between trials (1) and (3) there
is not such a significant change in mean as with the other trials.

~: At the time of this investigation, the pressure under which

the pump operates in general production had not been determined.
Consequently the pressures applied are somewhat lower than in actual
production conditions.
 145.
APPENDIX V
TEST FOR NORMALITY - NOISTUR:.<; V.ALUES

Grand Mean= 15.5g4

Standard Deviation = 0.173

Class I X-µ I X
~
Normal
Area Expected Observed 2
Ne an IZ= Prob. Frequency Frequency
I ~

15.20 -0. 3841 -2.26 0.0119 o. 011 9 2. 31 2

1 5.25 -0.334 -1. 96 0 . 0250 0 0131
0 2.54 5 o.s42
15°30 -0.284 -1. 67 0.0475 0.0225 4.37 5
15.35 -0. 234 -1.38 0.0838 o. 0363 7.04 8 0.1303
15.40 -0.184 -1 .os o. 1401 0.0563 1o.922 15 1. 523
1 5.45 -0.134 -0.79 0.2148 0.0747 14. 49 19 1 .402
15.50 -0.084 -0.49 0.3121 0.0973 18.88 22 0.517
15.55 -0°034 -0.20 0.4207 o. 1 086 21 • 07 29 2.986
15.60 +0.0161 +0.09 0 . 5359 0.11 52 22.35 12 4. 791
15.65 0.0561 0.39 o. 6517 0.1158 22.47 16 1 .860
15.70 o. 11 6 o.68 o. 751 7 0.1000 19 . 40 12 2.822
15.75 0.166 0.93 0 . 8365 0.0848 16.45 22 1 . s 72
1 5.80 0.21 s 1.27 0 . 8980 0.0615 11 93
0 11 0.073
1 5.85 0.266 1. 56 0.9406 0.0426 8 .2 6 5 1.289
1 5. 90 o. 31 6 1. 86 o. 9686 o . 02so 5.43 9 5.432
15.95 0.366 2.09 0.9817 0 . 0131 2 .54 0 2.743
16.00 0.4161 2.45 0.9929 0 . 0112 2. 17 0
16. 05 0.466 : 2.74 0.9969 0. 0071 1.38 2
25.20

Number of cells = 14
Degrees of freedom 11

~ 2 o. 01 = 24.73
2
$ o. 005 26.76

2
Since the calculated ~ value is within the acceptable limits,
the frequency distribution of Moisture Values can be said to approximate
to a normal distribution.
 146.

TEST FOR NOK1ALITY - SALT VALUES

Grand Mean -- 1. 57
Standard Deviation = 0.11

Class No r mal Expe ct ed Ob served

fTean
1
z = V.-µ Prob .
Area
Frequency Frequency Q
,2

1. 20 - 3. 391 o. 0003 0. 0003 0. 06

1.25 - 2. 936 0. 0016 0. 0013 0. 25
1. 30 - 2. 482 0. 0066 0.0050 0. 97 0 . 20
1. 35 - 2. 027 o. 0212 0. 0146 2. 8 3 3
1.40 -1°573 o. 05s2 0.0370 7.1 s 12 3. 25
1.45 --1.1 81 0 11 90
0 0. 0600 11 79
0 16 1. 50
1.50 - 0. 664 0 . 2546 0 .1 356 26 . 30 33 1. 71
1 • 55 - 0. 209 0. 4168 0.1 622 31 . 46 37 0. 98
1. 60 +0 . 245 0.5937 0. 1819 35 . 28 28 1.49
1. 65 +c . 700 0. 7580 0 .1 593 30. 90 26 0.78
1.70 1 . 1 54 o . s749 0. 11 69 22 . 67 24 o. os
1.75 1. 609 o . 9463 0. 071 4 13. 85 5 5.6 6
1. 30 2. 064 o. 9803 0. 0340 6 59
0 6 0. 05
1. 85 2.573 o. 9941 o. 013s 2 . 68 1. 07
1. 90 2. 973 0 . 9955 0. 0059 1 • 15 0
1. 95 3. 881 0. 0015 0. 3
16.77

Number of cells 11
Degrees of freedom = 11 -3
8

~ 2 o. 05 = 1 5 51 0

$ 2 o. 025 = 17. 54

Since the calcula ted $ 2 va l ue is within the acceptable limits ,

the frequency distribution of sal t values can be said to approximate
to a normal distribution .
 14 7.
APPENDIX VI

TABLE XVI
SUflII-1.Ar1.Y OF RESULTS 1''0R VARIABLES OTHER
THAN MOISTUIIB OR SALT

Sample Mean Standard

Date Variab l e
Size Value Deviation

7/2/72 Cream F'at Content 40 39.5 0 8

3/3/72 Cream I•'at Content 40 39 . 5 o. 6
6/4/72 Cream Fat Content 40 39 . 6 0. 5
Ribbon Speed 1 037 271 6
0 14- 4
0

3/5/72 Ribbon Sp eed 624 280 . 4 7. 3

1 6/5/72 Ribbon Speed 1 005 257.3 6.8
Salt Pump Press1ITe 2124 85. 1 5. 3
22/5/72 Ribbon Speed 1040 323 . 1 23 . 3
Sa l t Pump Pr essure 1 61 9 78 . 6 5. 0
26/5/72 Ri bbon Speed 1040 33s .2 18. 0
Salt Pump Pressure 1760 92 1
0 7.3

405
6/4/72 Ribbon Speed ( 1- 404) 25s . 9 12.8
633
6/4/72 ' Ribbon Speed (405- 279 . 7 s .7
I 1 03'/)
.

Unit s: Fat Content (%)

Ribbon Speed pulses/uni t t i me
Pressure (undefi ned)
 1480
APPEND I X VII
LISTING OF S~LBCTED cm~PU'rBR PROGRAMS
USED I N DATA PROcz; ssING

Program L Analysis of slices ~~d time se quenced seri es , moisture

and salt va l uesc

Data I nput : Ea ch pair of duplicate moi sture and salt values

we r e punched on one car d.
Iioisture va l ue - For mat F 7 . 4
Salt va lue - Format F 5. 3

Program 2 Ca lcul at ion of standard devi at ion and correlation

coeffi ci ent, displ acement anal ysi s included.

Data I nput: The mean mo i stur e and salt va lues of duplicate

anal yses we r e punched on one car d·,
Tfo i stur e v a lue - Format F 7. 4
Sa lt va lue - Format F 5. 3

Program 3 Ribbon speed and salt pump inj ecti on pr essur e data and
and cusum pl otted, Each dat a value pl otted is me211 of 8
consecutive r aw data values.

Dat a Iny.rut :
Ribbon Speed ·- For mat F 3 . 0
Sal t Pump Pressure - For mat F 2 . 0
For s a l t ~0ump pressure, va lue s gr eater than 1 00 were punched as
'units ' and 'tens ' digits on ly , The pr ogram r es tored the
'hundreds' digit.
 149

PROGRAM FOR ANALYSIS OF SLIC E S AND TIME SEQUE NC ED SE RI E S

DIMENSIONX(20 , 20),S(20t20)
XS UM=O •
XSQ=O •
SSUM=O •
SSQ=O •
XNA=O .
XNB=O •
XTOT=O •
XMEAN=O •
SMEAN=O .
SDEVX=O •
SDEVS=O •
SDEVY=O •
SDEVT=O .
YMEAN=O •
TMEAN=O•
ZMEAN=O •
UMEAN=O .
SDEVZ=O •
SDEVU = O •
READlOO , NA , NB
DOll = l t NA
DOlJ=l , NB , 2
1 READ200 , x ( I,J) . s< I ,J) . x< 1. J+l) .sc I ,J+l)
D02I=l , NA
D02J=l,NB
XSUM=XSUM+X(l,J)
XSQ=XSQ+X(l ,J )**2
SSUM=SSUM+S ( l,J)
2 SSQ=SSQ+S(l ,J) ** 2
XNA=NA
XNB =NB
 150

XTOf =XNA*XN B
XMEAN=XSU M/XTOT
SMEAN=SSUM/XTOT
SOEVX=SQRTF((XTOT*XSQ - XSUM**2)/ ( XTOT*( XTOT-1.)))
SOEVS=SQRTF (C XTOT*SSQ - SSUM**2)/ ( XTO T *CX TOT-l .)))
PR I NT700
PR INT 750
PRINT300,XMEAN,SDEVXtSMEANtSDEVS
PRINT800
PR I NT 750
0041=1 , NA
YS UM=O •
YSQ=O•
TS UM=O •
TSQ=O•
003.J=ltNB
YSUM=YSUM+X(l,J)
YS Q=YSO+X (I• .J) **2
TSUM=TSUM+S(l,J)
3 TSQ=TSO+SClt..1)**2
YMEAN=YSUM/XNB
T ME AN= TSU M/XNB
SOEVY=SQRTF((XNB*YSQ-YSUM**2)/(XNB*(XNB-1 .) ))
SOEVT=SQR TF( ( XNB*TSQ-TSUM**2)/(XNB*CXNB-l.) ))
PRINT300,YMEANtSDEVY,TMEANtSOEVT
4 CONTINUE
PRINT900
900 FORMAT ( 1Hl,40X,26HRESULTS FOR LIKE LOCATIONS//)
PRINT 750
006.J=ltNB t 2
ZSUM=O .
ZS Q=O •
USUM=O.
USQ=O •
 151

D051=1,NA
ZSUM=ZSUM+X(J,J)+X(I,J+l)
ZS Q=ZSQ+X (I, J) ** 2+ X (I, J+ 1) **2
USUM=USUM+S(I,J)+S(I,J+l)
5 USQ=USQ+S(I,J)**2+S(l,J+l)**2
XNAB=XNA*2•
ZME AN=ZSU M/XNAB
UMEAN=USUM/XNAB
SDEVZ=SQRTF((XNAB*ZSQ-ZSUM**2)/{XNAB*( XNAB-le)))
SDEVU=SQRTF((XNAB*USQ-USUM**2)/(XNAB*(XNAB-l.)))
PRINT300,ZMEAN,SDEVZ,UMEAN,SDEVU
6 CONTINUE
1 00 FORMAT (212)
200 FORMAT(lOX,F7.0,F5.0,F7eO,F5.0)
300 FORM~(lHo,2ox,F10.5,13x,F10.5,14x,F10.5,13x,F10.5)
700 FORMAT(1Hlt40Xtl5HOVERALL RESULTS//)
750 FORMAT(lH d5Xd4HMOISTURE MEAN ,8Xt17HMOISTURE STD DEV ,12X,
llOHSALT MEAN ,ax,13HSALT STD DEV///)
800 FORMAT(1Hl,40X,29HRESULTS FOR INDIVIDUAL SLICES//)
END
 152

PROGRAM FOR CALCULATION OF STANDARD DEVIATION AND CORRELA TIO N

COEFF ICIENT- DISPLACEMENT ANALYSIS INCLUDED

DIMENSIONX (40), S (40),F(40)

1 READ l OO , N
1 00 FORMAT (I 2)
003 I= 1, N
3 READ200 , X (l), S(l ), F(I )
200 FORMAT(l OX , F7 . 0 , F5 . 0,F4 . 0)
XS UM=O •
XSQ=O .
SS UM=O •
SSQ=O •
FS UM=O •
FSQ=O •
XPROD=O •
SDEVF =O.
SDEVS=O •
SDEV X=O •
FPROD=O•
R=O •
XN=O .
M=O
D041= 1, N
XSUM=X (I)+ XSUM
XSQ=X( I ) **2+XSQ
SSUM=S (I)+ SSUM
SS Q=S (I) **2+ ssa
FSUM=F( I )+ FSUM
FSQ=F CI ) **2+F SQ
4 XPROD=X (I) *S(l )+XPROD
XN=N
NT=N
FMEAN=FSU M/ XN
 153

SMEAN=SSUM/XN
XME AN=XSU M/XN
SDEVX=SQRTF((XN*XSQ-XSUM**2)/(XN*(XN-le)))
SDEVF=SQRTF((XN*FSQ-FSUM**2)/(XN*CXN-le )))
SDEVS=SQRTF((XN*SSQ-SSUM**2)/(XN*(XN-le)))
R=(XN*XPROD-(XSUM*SSUM))/SQRTF((XN*XSQ-XSUM**2)*(XN*SSQ-SSUM**2))
PRINT 300
300 FORMAT ( lH t40Xt24HCORRELATION COEFFICIENTS///)
PRINT400tR
400 FORMAT(lH t20Xt13HM:JISTURE-SALTtF8e5////)
PRINT130
130 FORMAT(lH t5Xt13HMOISTURE MEAN,5Xt16HMOISTURE STD DEV,8Xt9HSALT ME2
2ANt5Xt12HSALT STD DEV,8Xt8HFAT MEANt5XtllHFAT STD DEV///)
PRINT150tXMEANtSDEVXtSMEANtSDEVS,FMEANtSDEVF
150 FORMAT(lH ,ax,F10.s,ax,F10.s,11x,F10.s,6XtFl0.5,8X,Fl0.3,5X,Fl0.3/3
3 //)

SECTION FOR DISPLACEMENT ANALYSIS

PRINT500
500 FORMAT(lH t25Xt19HDISPLACEMENT VALUES,24X,14HCORRELN COEFFT//)
5 FSUM=O.
FSQ=O•
XSUM=O •
XSQ=O•
FPROD=O•
D06I=ltN
XS UM=X (I ) +XSUM
XSQ=X(l)**2+XSQ
FS UM=F ( I ) +FSUM
FSQ=F (I ) ** 2+F SQ
6 FPRDD=F (I )*X (I )+FPRJD
XN=N
R = ( XN*FPROO-( X SUM*F SUM) ) /S QRTF ( ( XN* XSQ-XSUM* *2) * ( XN*F5 Q-FSUM **2) )
M=Nf-N
 l{)

•
Cl)
u_
•
x
(jl
f't')
•
.;:t
......
• 0
x .....
fl') .
~
..
er f't') 0
• ....... .....
0
0
...
• I I'-
0 ..
+
z ..... ....f-
x
f- ..... x w
'z°
f- <(
~
fl')
I
::E:
.....
I
II
..... II l{)
0 _J
...... er z Cl) f- _J a
er u_ II 0 0 z
Cl. IL
0
.... z a x l?
<(
u w
0 I'- Cl) 0
0 .....
'°
 155

PROGRAM FOR RIBBON SPEED AND INJECTION PRESSURE DATA AND CUSUM
PLOf, USING MEAN OF 8 CONSECUTIVE VALUES

DIMENSIONSPC26),PR(40),SMN(220),PMN(220),CUSUM(220),CUSUMP(220)
50 READ40,MS,MP,XM
40 FORMAT(212,F4.0)
PR I NT 4 1 , MS ,M P , XM
41 FORMAT(lH 3HMS=,J4,lOX,3HMP=,I4,lOX,3HXM=,F8.0,//)
L=O
D08K= l t MS
READlOO,(SP(l),l=l,26)
100 FORMAT(26F3.0)
D08I=lt5
J= ( ( I -1 ) *5 ) + 1
S=CSP(J)+SP(J+l)+SP(J+2)+SP(J+3)+SP(J+4))/5e
L=L+ 1
SMN(L)=S
8 CONTINUE
PR I NT 200
200 FORMAT(lH ,2ox,22HMEAN OF 5 VALUES SPEED//)
PR I NT 2 1 0 , ( S"1 N ( I ) , I =1 , L )
210 FORMAT(15F8.0)
M=O
D028K=l ,MP
READllO,(PR(I),1=1,40)
110 FORMAT(40F2.0)
D0261=1,40
IF(PR(I)-13.)24,24,26
24 PR(l)=PR(I)+lOOe
26 CONT I NUE
00281= 1,5
J=(( 1-1 )*8)+1
P=(PR(J)+PR(J+l)+PR(J+2)+PR(J+3)+PR(J+4)+PR(J+5)+PR(J+6)+PR(J+7))/
18.
 156

M=M+l
PMN(M)=P
28 CONTINUE
PRINT220
220 FORMAT(1Hlt20Xt25HMEAN OF 8 V.<\LUES PRESSURE//)
PRINT230, (PMN ( I ) , I= l tM)
230 FORMAT C 15F8.0)
SECTION FOR CUSUM CALCULATION

PRSUM=O •
SPSUM=O.
COSUM=O •
DI FF=O •
XLf=L
PRINT400
400 FORMATC1Hlt30Xt31HSPEED CUSUM VALUES (MEAN OF 8 ) / / )
410 FORMATC1Hlt30Xt31HPRESS CUSUM VALUES (MEAN OF 8 ) / / )
00411I=1, L
SPSUM=SPSUM+SMN(I)
411 PRSUM=PRSUM+PMN ( I )
SP MUN= SPSU M/XL T
PRMUN=PR SUM /XLT
004121=1 t L
DIFF=SMN(l)-SPMUN
COSUM=C OSU M+DI FF
412 CUSUM(l)=COSUM
PRINT500, (CUSUM( I) , I = l , L )
500 FORMAT(lH t15F8.0)
PR I NT4 10
COSUM=O •
004131=1,L
DIF~=PMN(l)-PRMUN

COSUM=C OSU M+D I FF

413 CUSUMPCI >=COSUM
PRINT500, (CUSUMP(I) t l = l ,L)
 157

SECTION FOR PLOTTING OF DATA AND CUSUM VALUES (MEAN OF 8)

XMIN=l•
XMAX=XLT
XL=20 •
XO= 10 •
YL=lO.
YD= 10 •
YM AX=4 00 •
YM IN= 200 •
CALLPLOT( l.XMIN.XMAX.xL.xD.YMINtYMAX.YL,YD)
D0414I=ltL
X=I
CALLPLOT(O,X,SMN(I))
414 CONTINUE
CALLPLOT(7)
XMIN=l•
XMAX=XL T
XL =20 •
XD=lO.
YL= 10 •
YD= 10 •
YMAX=l20.
YMIN=O•
CAL LPL OT ( 1 t X MIN, XM AX tX Lt XO, YMt N t YMA Xt Y L, YD)
D0415I=l•L
X=I
CALLPLOT{O,X,PMN{I))
415 CONTINUE
CALLPLOT{7)
XMIN=l•
XMAX=XLT
XL=20 •
XD= 1 0 •
 158

YL= 10 •
YD=2 0 •
ACCEPT , YMAX, YM IN
CALLPLOTCltXMINtXMAX,XLtXDtYMIN,YMAX,YL,YD)
D0416I=ltL
X=I
CALLPLOT(OtXtCUSUM(I))
4 1 6 C 0 NT I NU E
CALLPLOT(7)
XMIN=l•
XMAX=XL T
XL=20 •
XD=lO.
YL=lO.
YD=20.
ACCEPT,YMAX,YMIN
CALLPLOT(l,XMIN,XMAX,XL,XDtYMIN,YMAX,YL,YD)
D04l7I=1, L
X=I
CALLPLOT(O,x,cusuMP(I))
417 CONTINUE
CALLPLOT(7)
GOT050
418 CALL EXIT
END

- - - - - - - - - -- - -
 159.
BIBLIOGRAPHY
SECTION A

(1) Preliminary Report on Trial of Contimab Continuous Buttermaking

Machine, October 1964 - April 1965.
Dolby, Jebson, Le Heron
Dairy Research Institute (N.Z.) Project B 145·

(2) Report on Westfalia Buttermaking Machine (February 1 967).

Dolby
Dairy Research Institute (N.Z.) Project BF 29 .

(3) Trials of Silkeborg Continuous Butterma.king Machine.

R.T. Anderson - Rangitaiki Plains Co-operative Dairy Company,
Edgecumbe, (N.Z.)
New Zealand Journal of Dairy Technology Vol. 4 page 29.

(4) Moisture in Butter in Relation to Dielectric Measurements.

Parkash and Armstrong
Journal of Dairy Science Vol. 52 , No. 8, page 1224·

(5) Measurement of Dielectric Constant of Butter.

Parkash and Armstrong
Dairy Industries, October 1970 Vol. 35, No. 10, page 688-9.

(6) Dielectric properties of butter and their application for measuring

moisture content during continuous processing.
Sone, Taneya, Randa .
XVIII International Dairy Congress (1970) Vol. 1 E, page 221.

(7) Continuous Estimation of Moisture in Fritz Butter.

Koenen (1968)
Die Molkerei - Zeitung 'Welt Der Milch' Vol. 22, No. 38,
page 1257-60.

(8) Ak:vameter MA-1 for Measuring Moisture Content in Continuous

Butterma.king.
Fexa and Rosenbaum (1965)
Prumsyl Potravin Vol. 16, No . 6, page 281-6.
(Dairy Science Abstracts Vol. 27, No . 3917) .
 160.

(9) Akvameter MA-1 for Buttermaking

Cerna and Vedlich (1965)
Prumsyl Potravin Vol . 16, No. 6, page 287-89.
(Dairy Science Abstracts Vol. 27, No. 3917).

(10) Possibilities of Automatic Control of Moisture Content in the

4 MVCS Continuous Buttermaking Machine "
Fexa and Rosenbaum (1966)
XVII International Dairy Congress (1966) Vol. C, page 421-30·

(11) Automatic Regulation of Moisture Content of Butter.

Fexa (1967)
Prumsyl Potravin Vol. 18, Mo. 12 , page 61 9-23 .
(Dairy Science Abstracts Vol. 30, No. 1928).

(12) Determination of Water Content in Butter by Means of Czech Akvameter .

Dolezalek and Rosenbaum (1 970)
XVIII Int ernational Dairy Congress (1970) Vol . 1E, page 222.

(13) N.R.C. Develops Microwave Instrument to Measure Moisture Content

of Butter.
National Res ear ch Council of Canada
Science Dimension Vol. 2, No . 2, page 24.

(14) Continuous Measuring Techniques for Proces s Control.

Slight ( 1 970)
Food Manufacture November 1970.

(15) West German Patent Application 1910217

Measuring Capacitance for Paste Like Substances.

(16) Evaluating Errors of Moisture Meters .

Berliner (1968)
Measurement Techniques No. 4 (April 1 969.), page 551-4.

(17) Testing of the Permittivity Type Oil-Moisture Meters.

Melkum.yan, Klugman, Kovylov (1968)
Measurement Techniques No. 8 (Aug. 1969), page 1132-4.
 16L

(18) Selection of the Physical Characteristics and Instrument Transducer

for Continuous Process Testing.
Shapiro (1968)
Measurement Techniques No. 8 (Aug. 1969). page 1097-1101.

(19) Determination of Moisture in Dairy Products by Near Infra-Red

Absorption of Methanol Extracts .
Goulding and Manning (1969)
Journal of Dairy Research (1 970) Vol. 37, No. 1, page 107.

SECTION B

(Include Refs. 1, 2, 3)

(20) The Present Position of Continuous Churning in New Zealand.

Dolby
New Zealand Journal of Dairy Technology Vol. 1, page 66f.

(21 ) Continuous Buttermaking in New Zealand

Dolby ( 1 967)
Dairy Industries Vol. 32, No. 7, page 575·

(22) Developments of Continuous Buttermaking in New Zealand.

Dolby ( 1 970)
XVIII International Dairy Congress (1970) Vol. 1 E, page 218.

(23) Literature from the Continuous Buttermaking Operators Course

(Dairy Research Institute).
Edited - R. Russell .

(24) General Principles of Continuous Buttermaking.

Dolby ( 1 969)
Dairy Research Institute News Circular Vol. 24, No. 3·

(25) Some Thoughts on Continuous Buttermaking in New Zealand.

Russell
Dairy Research Institute Publication 488.
 162 .

SECTION C (Analytical Methods)

(26) Determination of Moisture i n Butter .

Butter maker s Manual - McDowall
Vol. 2, page 153s.

(27) Deter mination of Salt in Butter .

Buttermakers Manual - McDowall
Vol. 2, page 1542.

(28) Quant i tative Chemica l Ana l ysis·

Fischer and Peters page 375.

( 29) Determinati on of Fat in Cr eam (Babcock) .

Butter makers Manual - HcDowall
Vol. 2, page 1502.

(30) Determination of Acidity in Cream.

Butter makers Manual - McDowa ll
Vol. 2, page 1 529.

(31 ) Buttermak:ers Manual - McDowall

Vol. 2 , page 1534.

(32) Mathematical and Sta tistical Techniques for Industry -

Cumulative Sum Techniques.
I.c.r. Monograph No . 3.

(33) New Form Statistical Tables .

G. Fisher .