close

Вход

Забыли?

вход по аккаунту

?

Production Variance in Purified Carboxymethyl Cellulose (CMC) Manufacture.

код для вставкиСкачать
Dev. Chem. Eng. Mineral Process., 12(1/2), pp. 217-231, 2004.
Production Variance in Purified
Carboxymethyl Cellulose (CMC)
Manufacture
Veronica Stigsson", David I. Wilson and Ulf Germgird
Dept of Chemical Engineering, Karlstad University,
Karlstad, SE 651 88 Sweden
This study investigates the variance in industrial production of carboxymethyl
cellulose (CMC). Initially it was thought that the product exhibited excessive quality
variations. However careful viscosity and molecular weight distribution
measurements on both the raw material and the final product, supplemented with
online quality measurements, showed that the product variations were less severe
than originally thought. The most probable reason for this unfounded quality concern
is due to imperfect routine measurements coupled with a limited reporting of the
subsequent ofline laboratov analysis. The situation where the routine quality
analysis plays as much a role in the uncertainty as the production itself is believed to
be of general relevance for the chemical processing industries.
Introduction
Excessive product variance in chemical manufacture stems either from raw material
variations, production upsets, or uncertainties in the quality analysis. While the latter
source is arguably the most benign, it is, in the authors' experience, more common
than is typically realised and it is believed that the consequences are of general
interest to those involved in the chemical processing industries. This hypothesis is
supported by taking the production of carboxyrnethyl cellulose (CMC) as an industrial
case study. The investigation into the operation of CMC production originated from
the concern that some grades of CMC exhbited excessive quality variations,
particularly for viscosity. Prior to this study, the engineering staff hypothesized that
excessive temperature and pressure deviations both within a batch and between
reactors contributed to the quality excursions. The goals of the study were to
establish the probable causes for the large variations in viscosity for some grades
(whde not seen in others), suggest possible remedies, and to obtain a better
understanding for the mechanism of CMC production. We describe the important
characteristics of CMC, how it is produced, and the plant trials aimed to quantify the
production. The results are summarised, and certain factors which are believed to be
generally applicable across many industrial chemical plants are highlighted.
* Author for correspondence (vcronica.stigsson@kau.sc).
217
K Stigsson, D.1 Wilson and LI. Germgdrd
Commercially Exploitable Properties of CMC
Carboxymethyl cellulose (CMC) is an anionic ether of cellulose and its major
commercial derivative. Powdered purified CMC is an extremely biologically and
chemically inert water-soluble form of cellulose making it suitable as an additive in
pharmaceutical and food products. Since CMC has a high molecular weight of
between 5 x lo4 and lo6, it has the ability to form weak three-dimensional gel
structures that retain water molecules with the assistance of hydrogen bonding. This
property modifies the flow behaviour of aqueous suspensions and solutions over a
wide range of viscosities and is the main commercially exploitable property of CMC,
[ 1,2]. A historical background to CMC production, tracing the development from the
1920s to the present day, has been previously reported [3], and the main chemical
properties are available [4].
There are at least 200 different standard grades of CMC. They are used in such
diverse applications as thickeners and stabilizers in food products, pharmaceuticals,
detergents, toothpaste, oil drilling mud, and paper coating. CMC can be produced
either as a crude technical product containing 25-40% by-product of sodium chloride
and sodium glycolate salts, or as purified product. These refined grades are washed to
contain less than 2% salts, or in the case of the extra-pure grades, less than 0.5% salt.
Reological Properties of CMC
The primary motivation to use CMC is to modify the viscosity of an aqueous solution.
While the viscosity is proportional to the average chain length of the CMC molecule,
the flow properties are also influenced by the molecular weight distribution (MWD),
the degree of substitution, and the evenness of the substitution, [ 5 , 61. For some
grades of CMC such as that used in ketchup, it is only the point viscosity at a given
strain rate that is required by the customer, while for other applications, the
rheological specifications are more complex. For additives to dairy products, and
particularly sauces and dressings, a shear thinning behaviour is desirable. For
toothpaste, a time dependence in viscosity is required. As additives to some food
products, the product viscosity must hit two target viscosity points at different strain
rates. The various different combinations of viscosity characteristics that are
commercially important are illustrated in Figure 1.
A theoretical study of CMC would ideally enable one to derive any properties of
interest (such as viscosity and its time dependence) from the molecular structure. To
make these predictions we would expect to know at the very least, the average
molecular weight, the molecular weight distribution, degree of substitution, the
evenness of the substitution along the chain, and gel content of the CMC. It is a longchain polymer which has undergone substantial treatment, therefore these variables
will be statistical quantities. Furthermore reliably determining either the molecular
weight distribution or the evenness of substitution remains a research topic, [7-91.
While the state properties of the polymer (MWD, evenness, etc.) are unknown, the
derived variables such as viscosity, moisture content, gel content, and degree of
substitution are relatively easily measured in the laboratory using standard techniques.
Furthermore, the moisture and gel contents can be inferred online using an NIR
instrument.
218
Production Variance in Purified CMC Manufacture
highly time dependent
e.g. toothpaste
e.g. ketchup
1
strain rate, y
strain rate,
24
time [hr]
Y
Figure 1. Alternative rheological requirementsfor
CMC.
The most convenient CMC viscosity analysis uses a traditional Brookfield
viscometer at a strain rate of approx. 3 s-'. While this technique is fast, economic, and
appropriate for production control, it gives no indication of strain rate dependence, or
thixotropic behaviour. The Haake rheometer addresses these latter deficiencies, but
has the disadvantage that it is not sufficiently robust for routine production analysis.
Plant Process Flow Diagram
A simplified process flow diagram for the industrial production of purified CMC is
shown in Figure 2. The utilities (steam, vacuum, cooling water and alcohol recovery)
are crucial for economic operation, but are not shown in detail in Figure 2.
NaOH. MCA
Washing filter
Alcohol
,...._____..__..__
.........
...........
Batch Reactors
Cellulose
!.I.
to recovety
Grinder
Silos
Slurry preparation
j ovcrsizc
.!
Stripper & Dryer
CMC gnnder
packmglng
Figure 2. The process flow diagram for purified CMC production.
A brief description of the manufacturing process is as follows. The raw cellulose is
fed via a buffer storage vessel to multiple batch reactors. In a reactor, the hydroxyl
groups on the C-2, C-3 or C-6 positions in the anhydroglucose ring become ionised by
reacting with NaOH in a mercerisation step:
25°C
Cellulose + NaOH -b
Cell-0-Na' + H20
219
V. Stigsson, D.I. Wilson and U.Germgdrd
The mercerisation is immediately followed by an etherification reaction in the
same vessel where the alkali cellulose is reacted with monochloroacetic acid (MCA),
or sodium monochloroacetate to yield CMC:
Cell-O--Na'
+ ClCH2COOH + NaOH
70°C
Cell-0 - CH2COONa + NaCl + HzO
However part of the sodium hydroxide simultaneously reacts with the
monochloroacetic acid in a side reaction:
HOCHzCOONa + NaCl + H 2 0
2NaOH + CICHzCOOH
Variable amounts of salts (sodium chloride and sodium glycolate) are produced, and
are subsequently removed in the purified CMC production. The average number of
groups substituted per cellulose ring (degree of substitution, DS) is theoretically
between 0 and 3. For commercial grades, typically between 0.5-1.5, [ref. 10, p.2211.
Figure 3 shows a typical reaction temperature profile starting with a mercerisation
step at a relatively low temperature, followed by an etherification step at a higher
temperature. The batch time scale in Figure 3 and subsequent plots has been
normalised for proprietary reasons. By adjusting the process conditions, a wide range
of CMC grades with different characteristics may be obtained.
The batch reactors are operated out of phase and the product is emptied into buffer
tanks that dampen out flow fluctuations prior to a continuous alcohol-washing filter.
Then alcohol-saturated slurry is dried, milled and sieved to obtain an even particle
size distribution. The over-sized particles are recycled to just before the milling stage.
I
1
I
I
I
I
I
Time [Normalised units]
Figure 3. Reaction temperature profile for the mercerisation and
etherlfication reactions to produce CMC (see also Figure 4).
220
Production Variance in Purified CMC Manufacture
Industrial Plant Trials to Quantify the Production Variance
Consider the production variance for two representative grades of CMC, namely
CMC H used as a food additive with a relatively high nominal molecular weight of
lo6, and CMC M with a medium nominal molecular weight of 2 x lo5 used for
toothpaste. CMC H was selected because it is easily manufactured, with little
production variation. Conversely CMC M is challenging to manufacture, and at times
exhibits excessive variations.
The most plausible sources of product variations are due to variations in the
characteristics of the raw cellulose, differences in operation between the multiple
reactors, variations in the temperature program and the addition of catalyst, poor
mixing in the buffer storage vessels, incomplete washing and dqmg, and poor milling
operations. Superimposed are Variations due to the uncertainties in sampling and the
laboratory analysis. An experimental program was devised to quantitatively assess
these sources.
Section (i) following illustrates the nominal variations in the temperature profile
for two of the reactors, for both products. Section (ii) justifies the plug flow
assumption through the washing and drying stages necessary due to the sampling
constraints, and section (iii) analyses the product viscosity within reactors, across
reactors, and throughout the downstream processing stages. Section (iv) investigates
the possibility of using hysteresis to predict viscosity/time dependence and laboratory
analysis variations, while section (v) considers raw material variations.
(i) Reactor temperature variations
The temperature profile for two different reactors during production of CMC M is
shown in Figure 4(a) giving an indication of the repeatability of the batches within
one reactor, and the differences between reactors. Figure 4(b) shows the
superimposed reactor temperatures for five batches of CMC H from 2 reactors. The
insert plot in Figure 4(b) shows a zoomed portion around the etherification step
highlighting that reactor no.1 is consistently about 2-3°C above reactor no.2.
In the case of CMC M, the relatively high temperature for the mercerisation step,
(45"C), combined with a short contact time means that the reaction occurs only on the
outer exposed surface of the fibers deliberately ensuring an uneven distribution of
substituents. Conversely, the lower temperatures for the CMC H production are
intended to produce as evenly distributed product as possible. In the subsequent
etherification step, increasing the temperature favours the main reaction thereby
increasing yield and minimising salt production, but has little noticeable effect on the
evenness.
(ii) Plug flow assumption
Due to the aggressive nature of the reactants, sample taking was not possible before
the milling stage (shown in Figure 2). Relating the samples precisely to changes in
production required measurement of the transport delay, and quantification of the
mixing (if any) fiom the buffer storage to the first sampling point. Figure 5 shows the
221
V. Stigsson, D.I.Wilson and L! Germgdrd
--
I
I
c
(a) Temperature measurements in two reactors for CMC M. Solid line reactor
no. I , dotted line reactor no.2.
Reactor temperature
80
70
60
50
40
30
20
I"
~~
lime [Normalised units]
Temperature measurements (superimposed) in two reactors for CMC H.
Solid lines reactor no.I . dotted lines reactor no.2.
Figure 4. Reactor temperature comparisons.
222
Production Variance in Purtfied CMC Manufacture
]
]
f
f
1
O
on
0
5
10
15
20
25
30
35
40
45
50
55
Figure 5. Salt concentration (measured as conductivity) at the mills (points 0)
compared to a fitted non-ideal flow model (-).
resultant salt concentration (upper plot) just before the mills after the alcohol washing
was turned off for 5 minutes (lower plot). From this tracer analysis, we can quantify
the extent of the non-ideal flow by fitting a least-squares dispersion model with time
delay (solid line in Figure 5). The dominant poles of the fitted model give an idea of
the residence time (2-5 minutes), and the time delay is the transport delay (38
minutes) [l 1, 121. The slight dispersion is probably mainly due to the draining of the
alcohol over the 1 minute conveying time on the washer.
The plug flow has both positive and negative consequences. The minimal
dispersion indicates that there will be little mixing between different production
grades, and that the production line does not need to be extensively flushed during a
grade change. The drawback is that there is little scope for damping of natural
fluctuations due to mixing w i t h a production campaign of a single grade. The latter
could however be addressed by including a buffer tank with a stirrer.
(iii) Spread within a single reactor and a single campaign
Despite the variations in the reactor temperature within a single reactor (see Figure 4),
the viscosity of the CMC/water product for both CMC M and CMC H shows no
significant variations (see Figure 6a). The differences between the reactors as shown
in Figure 6(b) also indicates that the clear temperature bias evident in Figure 4 has
little effect on the quality of the final product. Variations in viscosity at low strain
rates (< 0.1 s-I) are not considered significant, and are difficult to accurately
determine. The more convenient Brookfield measurements plotted on Figure 6(a) at
an arbitrary strain rate of 3 s" exhibit good repeatability, but at a slightly lower
viscosity.
223
V. Stigsson, D.I. Wilson and U Germgdrd
As expected, the factory grinding stage lowers the viscosity slightly as shown in
Figure 7, while laboratory grinding has no effect. Mechanisms have been postulated
[13] for the effect of mechanical action on cellulose (which is believed to behave
similarly to CMC), and indicate that the physical cutting of the fibers in the grinder
coupled with the increased local temperature lowers the degree of polymerisation, and
hence viscosity. At extreme temperatures over an extended period of time, crosslinking of the substituted groups form an insoluble gel. Cross-linking does not occur
in cellulose.
(iv) Viscosity/time dependence
For some materials such as toothpaste it is desirable to have a viscosity/time
dependency where the solution thickens over a period of 2 to 3 days. The addition of
a solvent such as sorbitol or glycerine to the aqueous CMC solution has a synergistic
effect which exaggerates the viscosity development. Measurements on pure aqueous
CMC solution do not show a viscosity-time dependence.
Figure 8(a) shows the viscosity development of a CMC M/sorbitol solution
measured both sporadically with a Brookfield viscometer (0)and more regularly with
a Haake viscometer (0).In the latter case, the solution was measured multiple times
over a period of a few minutes, approximately every hour. The first measured point in
each series (*) is almost always an outlier, and subsequently ignored. The solid line
in Figure 8(a) is a least-squares curve fit of the remaining Haake data, while the
dashed line is for the Brookfield data assuming a first-order dynamic response. The
less reliable Brookfield viscometer does show a significant viscosity increase, but this
is not evident fiom the Haake results.
The distribution of substituents is believed to affect the time dependent behaviour
of CMC, [2, 141. For molecules where the substitutions are evenly spread along the
chain, CMC exhibits low time dependency. For the cases where there are long
stretches of the chain essentially unsubstituted, the CMC is thought to exhibit a high
degree of time dependence. The difficulty is that it is impractical to measure the
evenness of substitution, and the viscosity-time dependency tests are time consuming.
A rapid analytical alternative is to measure the viscosity hysteresis behaviour (see
Figure 1) where it is postulated that the area between the two curves is related to the
evenness of substitution and hence the time dependence. Figure 6(b) at least does not
invalidate this hypothesis since it shows that the CMC M exhibits little hysteresis,
c o n f d n g the Haake results from Figure 6(a).
(v) Raw material variations
As previously stated, the molecular weight distribution (MWD) of the final CMC
product is a key quality parameter. Excessive variations in the product MWD could
either arise fiom inherent variations in the raw cellulose material, or be introduced
during production. Figure 9 compares the MWD of the feedstock cellulose with
MWD of the product CMC, as derived from different cellulose grades and analysed
by size-exclusion chromatography [8,9]. Cellulose ‘A’ is derived from cotton lint and
is known to have a tight MWD because cellulose from lint is essentially unprocessed
224
Production Variance in Purified CMC Manufacture
7
lo'
b
!
E
10'
'
1
P
B
loo.
"104
10.'
loo
10'
.
loo.
10.11
10'
lo'
10
'
10.'
strain rate IS.']
roo
Stnin rate
10'
I
10
'
IS-'^
@) Viscosity from diflerent reactors
(a) Viscosity profiles of diferent
batchesfi-om the same reactor in the
same campaign. The Brookj?eld
measurements
at
a
single
(approximate) strain are given for
comparison.
in the same campaign, (Reactor no. 1:
hollow symbols, reactor no. 2 filled
symbols).
Figure 6; CMC viscosity measurements as measured using a Haake rheometer.
lo2
,
t
1(
10' -
After Lab grinding
4
Ip
Q
v
loo -
lo-' I
1o4
1o-2
1oo
1o2
1o4
Strain rate [s-']
Figure 7. Viscosity variations of CMC H before (open symbols), after laboratory
grinding, filled symbols), and afterfactory grinding filled symbols).
225
V. Stigsson, D.I. Wilson and U.Germghrd
10'
1
I
0
5
10
15
Tlme [hwrs]
0 Bmdtfield
11
20
25
(a) The time dependence of CMC
Mhorbitol solution over 24 hours
measured by Haake (e)
and
Brooweld (0)viscometers.
1
L
10.'
1oo
1oz
10'
Sttain rate IS-']
(b) The hysteresis of CMC Mhorbitol
solution measured by a Haake
rheometer after 1 hour and 24 hours.
Figure 8. Viscosity of a CMC MXorbitol mixture used to simulate a toothpaste
formulation. (In neither case does the material exhibit any significant evidence of
viscosity-time dependence.)
and contains no hemicellulose. Conversely the wood cellulose pulps 'B' and 'C' have
undergone an aggressive sulfite processing step which damages the polymer chains,
and both exhlbit the small hump at the lower molecular weights due to the
hemicellulose fraction. Because high viscosity cellulose (from pine or spruce) is
extremely difficult to dissolve [15], the MWD analysis presented in Figure 9 should
only be interpreted qualitatively, particularly on the abscissa scale. The inability to
find a reliable standard for cellulose MWD analysis is not unusual, and many studies
[e.g. 81 are therefore forced to present the results not as molecular weight or degree of
polymerisation (DP), but as eluted volume from the chromatograph column. Note
that the MWD of the lint cellulose was not measured for this study.
By comparing the shape of both the prior and final MWDs, and noting that the
relative position of the small hump at the lower molecular weights due to the
hemicellulose fraction in the wood pulps does not change significantly in the product
CMC, it is not unreasonable to postulate that the production itself does not
significantly alter the initial distribution. This infers that the CMC production steps
are not further damaging the polymer chains and thereby increasing the variance in
viscosity. Current research is investigating the possibility of measuring both the
CMC and cellulose molecular weight distributions in the same column in order to
give an improved direct comparison.
226
Production Variance in Purijed CMC Manufacture
1
0.8
0.6
1
0.4
0.2
0
10"
10'
1010'
10'
Degree of polymerisation of cellulose
10'
-
-
-
1o6
Mol. weight of CMC
Figure 9.
The molecular weight distribution of (upper) feedstock cellulose from
CMC from 3 dtflerent
cellulose suppliers. Cellulose MWD data from [I 61.
wood and resulting (lower) CMC for one grade of
(vi) Online NIR measurements
One relatively new online analytical measurement is to use Near Infra-Red (NIR) for
moisture and the degree of substitution (DS). NIR has an advantage over IR since the
samples exhibit lower light absorbance (which means the NIR penetrates the sample
better), and that it is possible to use optical fibers and so multiplex the instrument
increasing the utilisation of the single instrument. Figure 10 provides a comparison
between moisture and degree of substitution as measured in the laboratory ( 0 points)
22 7
V. Stigsson, D.I. Wilson and U. Germgdrd
and that calculated using a partial least squares (PLS) model built from the NIR
measurements (0 points) from the material just prior to grinding. Ths data obtained
over 4 days shows that NIR can be used online for the determination of moisture and
DS, at least to the same resolution as the laboratory measurements. Further details of
the instrumentation, signal pre-processing and model development are available [ 171.
An example of the influence of the pre-processing required prior to multi-variate
model building intended for production diagnostics specifically for CMC has been
reported [ 181.
Q
*t
0
1
I
I
I
0.5
1
1.5
U
I
I
I
I
I
I
0
0
I
2
2.5
time [days]
Predicted
Reference
1
1
I
I
1
3
3.5
4
Figure 10. Predicted values (0) compared to reference values (0) for moisture (%)
and degree of substitution vs. time.
228
Production Variance in Purified CMC Manufacture
(vii) Production run statistics
Figure 11 shows that the actual annual production data of the more challenging CMC
M lie mainly within the quality limits externally imposed by the customer. The odd
outliers for the CMC viscosity are due to the startup batches of the production
campaign since the reactors were used for other grades during this period. The
overall production quality compares favorably with the benchmarking studies [ 191
which reports values of sigma performance of 2.5% and complaints of 3% of orders
as being typical for the chemical industry.
r
,
1.21
1
I
I
1
0
0
1
o.2
t
0'
I
I
I
I
I
I
I
1
I
6
Q
0'
Jan
I
APr
I
Jul
Production date
I
Oct
I
Jan
Figure 11. Production run statistics for CMC M for almost 1 year showing: (a) the
variation in viscosity of 2% CMC; (3) the variation in viscosity for a
toothpaste formulation with quality limits (--).
229
V. Stigsson, D.I. Wilson and U.Germgdrd
Conclusions
The results of the quality tests of CMC show that the production delivers a consistent
product for both CMC M and CMC H. This is due to the apparent insensitivity of the
reaction to minor changes in temperature and to non-ideal mixing, and the plug flow
nature of the subsequent downstream processing that rninimises the potential
intermixing of different production grades. The CMC maintains the expected
molecular weight distribution of the raw cellulose and this strengthens the argument
that the production itself produces little additional variation.
However, the danger of erroneously over-estimating the production variance
stemming from industrial viscometer readings probably explains why it had
previously been believed that there was excessive product variation, although there
was little reason to be concerned. The Brookfield spread of data shown in Figure 6(a)
underestimates the problem because the strain rate for those measured values is so
poorly defined.
The production run statistics presented in Figure 11 are ‘world class’ [terminology
in 191 as quantified by the percentage of off-specification product compared to typical
chemical production facilities. This naturally raises the question as to why the
production staff had previously believed that there was a production problem with
CMC M, when there is no experimental evidence of any problems. The most probable
explanation is due to a cultural memory effect held by the operating staff. They
remember operational difficulties in the past, currently observe minor temperature
deviations coupled with rushed on-site viscosity readings using a Brookfield
viscometer, but they are not regularly informed of the production quality assessment
undertaken on the product in the weeks after the production. Closing this ‘feedback
loop’ simply by better communication will almost certainly have the desired effect.
Acknowledgements
The authors are grateful to Niclas Anderson, Anders Anderson, Oliver Rupert and
Garan Kloow for fruitful discussions and constructive criticism, and to the personnel
at Noviant for their help and support. Thanks also to Tembec Inc. for providing some
of the data used in this study.
References
1.
2.
3.
4.
5.
6.
230
MetsB-Serla CMC Products. 1996. Technical information. MetsB-Serla Chemicals AB,(now Noviant
AB), Skoghall, Sweden.
Metsi-Serla CMC Products. 1997. CEKOL cellulose gum for toothpaste. Technical report, (now
Noviant AB, Sweden).
Stigsson, V., Kloow,G., and GermgArd, U. 2001. An historical overview of carboxymethyl cellulose
(CMC) production on an industrial scale. Paper Asia, 17(10), 16-21.
Kloow, G. 1981. CMC: Properties and industrial applications. Technical report, Billerud, Karlstad,
Sweden.
Huhtala, M. 1997. Viscosity Measurement of CMC. Technical. Report, Metsa Specialty Chemicals
OY (now Noviant OY), ABnekoski, Finland.
Kulicker, W.M., Kull, A.H., Kull, W., and Thielking, H. 1996. Characterisation of aqueous
carboxymethylcellulosesolutions in terms of their molecular structure and its influence on rheological
behaviour. Polymer, 37( 13), 2723-273 1.
Production Variance in Purified CMC Manufacture
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
Puls, J., Homer, S., Kruse. T., Saake, B., and Heinze, T. 1998. Enzyme-aided characterization of
carboxymethylcellulose with conventional and novel distribution of functional groups. Das Papier,
12,743-748.
Schult, T. 2000. Properties of acid sulfite cellulose for cellulose derivatives. PhD thesis, Norwegian
University of Science and Technology, Trondheim, Norway.
Rinaudo, M., Danhelka, J., and Milas, M. 1993. A new approach to characterising
carboxymethylcelluloses by size exclusion chromatography. Carbohydrate Polyrners, 21,l-5.
Klemm, D., Philipp, B., Heinze, T., Heinze, U., and Wagenknecht, W. 1998. Comprehensive
Cellulose Chemistry: Functionalization of Cellulose, Volume 2. Wiley-VCH Publishers.
Levenspiel, 0. 1962. Chemical Reaction Engineering. John Wiley & Sons, New York.
Ljung, L. 1999. System Identification: Theory for the User. 2”dEdn, Prentice-Hall, USA.
Fiber Cutting Theory. Technical report, Buckeye Technologies Inc., 1001 Tillman, P.O. Box 80407,
Memphis Tennessee, USA. Web adddress http://www.bkitech.cod.
Gelman, R.A. 1985. Characterisation of cellulose derivatives: Distribution of substituent groups along
the chain. In J.F. Kennedy, (3.0.Phillips, D.J. Wedlock, and P.A. Williams (eds), Cellulose and Its
Derivatives: Chemistry, Biochemistry and Applications, Chapter 25, pp.293-300. Ellis Horwood
Limited, UK.
Sji)holm, E., Gustafsson, K., Eriksson, B., Brown, W., and ColmsjB, A. 2000. Aggregation of
cellulose in lithium chloride/N,N-dimethylacetamide.Carbohydrate Polymers, 41(2), 153-161.
Anon. 2002. MWD of cellulose. Technical report, Tembec International, Canada. Personal
communication.
Anon. 2000. NIR: for on-line analysis of moisture in CMC. Technical report, Noviant Group,
Skoghall, Sweden.
Andersson, N., and Wilson, D. 2003. Orthogonal signal correction inside cross-validation improves
error estimation in PLS models. In Modelling identification & Control, 10-13 February, Innsbruck,
Austria. pp.45-50.
Benson, R. 2001. When process control “disappears”- You know it has arrived. 6th World Congress
of Chemical Engineering. 23-27 September, Melbourne, Australia.
Received: 10 November 2001; Accepted afer revision: 20 March 2003.
231
Документ
Категория
Без категории
Просмотров
0
Размер файла
693 Кб
Теги
production, CMC, carboxymethyl, variance, manufacture, purified, cellulose
1/--страниц
Пожаловаться на содержимое документа