Printer Friendly

Performance evaluation of an interlocking rotor internal mixer.

Performance evaluation of an interlocking rotor internal mixer

Introduction There are many different criteria by which the performance of an internal mixer can be assessed; and a substantial number of process variables which can be adjusted to influence performance. Even in a limited process capability study, the investigator is faced with a multi-factor, multi-response problem. Fortunately, there are techniques available which can resolve many of the difficulties. In this article, statistical (or factorial) experiment design and multiple regression analysis modeling are utilized. The influence of six process operation variables on the processability of a carbon black filled natural rubber compound and on mixer responses is quantified.

The "processability" of a rubber compound depends on its rheological response to the deformation conditions and temperatures in down-stream processes and on its curing characteristics. In any practical process, a rubber compound is subjected to a spectrum of deformation rates and temperatures. Consequently, to characterize the influence of mixing variables on processability, tests which measure rheological properties over a range of conditions are useful. In addition, where high temperature ([is greater than] 160 [degrees] C) curing processes are used, curemeters with a rapid sample warm-up time should be selected, to obtain discriminating data at practical cure temperatures.

The mixer used for the work is a computer controlled Francis Shaw K1 Intermix (5.51 chamber volume with NR2 rotors) equipped with automatic feed systems for fillers and oil injection. Rheological behavior was measured with a Negretti TMS Biconical Rotor Rheometer and cure characteristics with a Wallace Precision Cure Analyzer.


Experiment design and mixing

The rubber compounds selected for the investigation is listed in table 1. It was designed to invoke all the normal aspects of rubber mixing: mastication, dispersion, oil absorption and cure system distribution, while being tolerant of the wide range of mixing times and batch temperatures expected in the mixing experiment.

A multi-factor approach to experimentation was adopted, using factorial experiment design [refs. 1 and 2]. Six mixing variables were chosen for investigation, by reference to previous work [refs. 3 and 4]:

[T.sub.1]-circulating water temperature for the chamber ([degrees] C)

[T.sub.2]-circulating water temperature for the rotors ([degrees] C)

S-rotor speed (rev/min)

F-fill factor

M-unit work of mastication (MJ/[m.sup.3])

D-total unit work of mixing (MJ/[m.sup.3])

A conventional mixing technique was selected, in which the cycle started with the feeding of the rubber, zinc oxide, stearic acid, DCBS, PVI and antioxidant to the mixer. The addition of the carbon black, together with the sulphur, was then computer-initiated after the input of the unit work of mastication. The oil addition was standardized at 40 MJ/[m.sup.3] after the carbon black addition and a sweep (ram raise and lower) carried out at 120 MJ/[m.sup.3] after the carbon black addition, prior to the mix being discharged automatically onto a two-roll mill after the input of the total unit work of mixing. One pass through the mill rolls with a 5 mm nip setting was used for sheeting.

A center-composite rotatable experiment design was selected. It is built up from a half replicate two level and six variable factorial, giving [2.sup.5] = 32 mixing conditions, supplemented by 12 "star" points and 12 center point replicates. The design is optimized for fitting a second-degree polynomial model, in which the two level factorial points enable linear and second order interactive effects to be modeled, the star points enable quadratic curvature to be modeled and the center point replicates provide for estimates of run-to-run error and model adequacy. For purposes of model estimation by multivariable regression analysis and subsequent statistical analysis, independent variable (factor) values are translated into code values. Table 2 defines the relationships between real and coded variables. The coded values for star points are defined, for the half-replicate, by:

[+ or -]a = [2.sup.(k-1)/4] where k is the number of independent variables. In table 3 the experiment design is specified. The two level factorial points for the half replicate were selected by the identity:

[V.sub.1] x [V.sub.2] x [V.sub.3] x [V.sub.4] x [V.sub.5] x [V.sub.6] = 1

The multi-variable regression analysis was accomplished with a statistical computer package named GLIM (Generalized Linear Interactive Modeling), release 3 [ref. 5]. The resulting model takes the form: [Mathematical Expressions Omitted] where Y is the measured response, [B.sub.0] is the zero order constant (numerical average of all measured responses); [B.sub.1], [B.sub.2], etc. are the linear coefficients: [B.sub.11], [B.sub.22], etc. are the quadratic coefficients and [B.sub.12], [B.sub.13], etc. are the interaction coefficients. The factors [T.sub.1], [T.sub.2], etc. have their coded values.


Having selected unit work as the means of controlling the mixing cycle, the mixing time and batch temperature become dependent variables or responses.

After mixing, the rheological behavior of the batches was measured with a Negretti Automation TMS biconical rotor rheometer. The steady-state flow curve at 100 [degrees] C was obtained for shear rates in the range 0.1 to 100 1/s with the testing program shown in table 4. The cure characteristics of the batches were measured at 170 [degrees] C with a Wallace Precision Cure Analyzer.

Results and discussion The coefficients of Eqn. 1 for selected responses are presented in table 5. Inspection of the table will reveal those coefficients which are large enough to be of practical significance and the asterisks indicate coefficients which are statistically significant at the 95% confidence level.

Before proceeding to the contour graphs derived from the response equations, some comment on the coefficients in table 5 is worthwhile. The mixing time is dominated, as expected, by unit work, but the influence of unit work on the other responses, particularly the material property responses, is less pronounced, throwing some doubt on the utility of mixing energy as a means of mixer control. The most generally influential variable is rotor speed, mainly through its strong relationship with batch temperature.

Consequently, increasing rotor speed increases apparent viscosities, due to the rise in temperature removing the batch rapidly from conditions for efficient mastication by mechanical chain scission. The circulating water temperatures for chamber and rotors also exert a strong influence on batch temperature in the early stages of the mixing cycle and, hence, on mastication efficiency and batch viscosity. In the later stages of the mixing cycle, after the addition of the carbon black and sulphur, batch temperature is dominated by rotor speed. The cure characteristics, as measured by cure time and maximum curemeter torque, show a strong relationship with rotor speed but not with circulating water temperatures. The significant interaction of rotor speed and total unit work for the cure characteristics can be explained by reference to the influence of unit work on mixing time. The rotor speed - unit work interaction is a measure of the heat history experienced by the cure system during mixing.

The contour plots are presented with axes scaled in coded values, but the contour values are "real". The limits of the experimental region are defined by: [Mathematical Expression Omitted] The coefficients of the polynomial equation are optimized for the best fit within the experimental region and cannot be relied upon for extrapolation outside it. In the contour plots presented, the experimental region is defined by a circle of radius 1.73 (coded values).

Total mixing time, as a function of rotor speed and total unit work (figure 1), shows an almost linear dependence on both variables. Mixing time increases with unit work, as expected, but it also increases with rotor speed. This unexpected result is mirrored in the mastication time response and is attributed to the rubber feeding characteristics of the mixer. An initial observation was that the 50mm cubes of natural rubber used in the experiment could respond elastically at high deformation rates and were thus difficult to entrain around the rotors. The ram down time confirmed that the ingestion time increased with rotor speed. Batch temperature at discharge, as a function of rotor speed and total unit work (figure 2), shows that rotor speed dominates batch temperature by its influence on the energy "balance" of the system. At low rotor speeds and high unit work levels the mixer approaches thermal equilibrium, but batch temperature continues to rise with increasing unit work (and time) at high rotor speed.

Cure time (to 100% conversion), as a function of rotor speed and total unit work (figure 3), shows a large region of insensitivity to both these variables. The reduction in cure time at low rotor speed and unit work can be dismissed as an artifact of the response surface fitting, and the rapid reduction at high rotor speed and unit work attributed to accumulated heat history, as mentioned earlier.

Maximum curemeter torque, as a function of rotor speed and total unit work (figure 4), shows a maximum in the region of cure time insensitivity identified in figure 3. Comparisons of the coefficients in table 5 for minimum and maximum curemeter torque reveal that the response surfaces are substantially different, and that the former has much in common with the apparent viscosity responses. Consequently, the shape of the response surface must be attributed largely to the dependence of maximum torque (state-of-cure) on the influence of rotor speed on batch temperature.

Melting of some cure system additives aids reduction of their scale of segregation in the rubber compound during mixing, and thus maximizes their effectiveness. The accelerator DCBS and the retarder Santocure PVI have melting temperatures of 95 [degrees] C (approx.) and 90 [degrees] C, respectively, which, with reference to figure 2, support the hypothesis that their incomplete melting and dispersion are responsible for the lower state-of-cure observed at low rotor speeds.

Apparent viscosity at 0.1 1/s, as a function of circulating water temperature in the chamber, ram and door (WT1) and the rotors (WT2) (figure 5), indicates that mastication, not dispersive mixing of the carbon black, dominates viscosity reduction. The efficiency of mastication by mechanical chain scission depends on the stresses generated in the rubber [ref. 6]. At constant rotor speed the influence of WT1 and WT2 on batch temperature early in the mixing cycle determines the stresses which can be generated. The rotor temperature is shown to be more influential than that of the chamber, etc., but the effect of each is similar and additive.

Apparent viscosity at 0.1 1/s, as a function of WT1 and rotor speed (figure 6), shows the combined influence of strain rate and temperature on stress. However, the influence of rotor speed on the rate of batch temperature rise limits the benefits to be gained from increasing it. At each water temperature there is an optimum rotor speed for viscosity reduction, and, as water temperature is increased, the optimum speed decreases. A similar response surface can be drawn for WT2 and rotor speed.

Apparent viscosity at 100 1/s as a function of WT1 and WT2 (figure 7), shows a similar general shape of response surface as at 0.1 1/s, but with a substantially reduced sensitivity to WT1 and WT2. The reduction of sensitivity can be attributed to two sources. Mechanical chain scission operates on long molecules preferentially, due to their more limited mobility (or longer relaxation times) [ref. 7]. Similarly, the effect of long molecules on apparent viscosity is dominant at low shear rates [ref. 8].

As shear rate is increased, progressively shorter molecules begin to influence the flow behavior. Since the shorter molecules are less likely to have been modified in the mastication process, the effect of diminishing dependence of apparent viscosity on mixing conditions as shear rate is increased is observed. Also, in the apparent viscosity data presented, the molecular weight distribution effect is augmented by shear heating in the biconical rotor rheometer used for the flow measurements.

As shear rate, shear stress and the duration of shearing increase during the test, the sample temperature rises above the set (instrument) temperature. Samples producing a high stress will generate more heat and be subject to a greater temperature rise, and a greater temperature induced reduction of shear stress. Consequently, a stress - temperature compensation mechanism will reduce discrimination between samples at high shear rates and long times. The deviation of the response surface shape at high shear rates from that at 0.1 1/s could also be due to the shear heating effect in the rheometer. However, other work [ref. 9] has shown that for natural rubber compounds, the level of carbon black dispersion becomes influential on viscosity at high shear rates.

Conclusions Quantitative relationships for the dependence of selected mixed compound properties on mixing conditions have been presented which provide an insight into the performance of the interlocking rotor type internal mixer for carbon black filled natural rubber compounds. A number of points emerge from interpretation of the response surface models:

* Within the range used, the material property responses are insensitive to the unit work of mastication and the total unit work. Since mastication time and total mixing time are approximately linearly dependent on the two unit work variables, insensitivity to mixing time can also be inferred.

* The material property responses relate better to batch temperature than to mixing time or unit work. This can be attributed to the dependence of batch temperature on the rate of energy input, since the processes of mastication and carbon black dispersion are stress controlled.

* The increase in mixing time with increasing rotor speed is attributed to the elastic response of the natural rubber to high deformation rates preventing effective engagement with the rotors.

* Cure time is dominated by the influence of rotor speed on batch temperature, but shows a large region of insensitivity which can be exploited in the selection of practical mixing conditions.

* The reduction of the state-of-cure (as measured by maximum curemeter torque) at low rotor speeds is attributed to the influence of rotor speed on batch temperature. Minimum temperatures must be achieved to melt and distribute effectively some cure system additives. Both the accelerator and the retarder used have melting points (95 [degrees] C and 90 [degrees] C) which are in a range to support the hypothesis.

The interpretation of the results presented has been selective and there is considerably more information to be culled from further examination of the response surface models.

Acknowledgements The work reported in this article was undertaken as part of a program funded by the Science and Engineering Research Council of Great Britain in conjuction with BTR Ltd., Du Pont Ltd., Francis Shaw & Co. Ltd., Pirelli Ltd. and Schill and Seilacher (U.K.) Ltd. [Tabular Data 1 to 5 Omitted] [Figures 1 to 7 Omitted]

References [1]G.E.P. Box, W.G. Hunter and J.S. Hunter, Statistics for Experimenters, John Wiley & Sons, New York (1979). [2]G.C. Derringer, Rub. Chem. Tech., 3, 61, 377 (1988). [3]K.B. Basir and P.K. Freakley, Kaut. u Gummi Kunst, 35, 3, 377 (1982). [4]P.C. Ebell, Ph.D. thesis, Loughborough Univ. (1981). [5]R.J. Baker and J.A. Nelder, "GLIM" release 3, Numerical Algorithms Group, Oxford (1978). [6]M. Pike and W.F. Watson, J. Polym. Sci., 9, 229 (1952). [7]A.I. Medalia and J. Heckman, J. Coll. Inst. Sci., 36, 173 (1971). [8]W.W. Graessley, J. Chem. Phys., 47, 6, 1942 (1967). [9]K.B. Basir, Ph.D. thesis, Loughborough Univ. (1985).

J. Batchelor, P.K. Freakley, S.N. Ghafouri and D.W. Southwart, Institute of Polymer Technology and Materials Engineering, Loughborough University, U.K.
COPYRIGHT 1989 Lippincott & Peto, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Southwart, D.W.
Publication:Rubber World
Date:Jul 1, 1989
Previous Article:Physical Testing of Rubber.
Next Article:On-stream oil purification lowers operating expenses.

Related Articles
Increasing output and quality by updating the mixing room equipment.
Skinner Engine.
Review of processing equipment.
Two-Stage Machine Promises Better Mixing.
Rebuilding in-house capabilities.
Laboratory mixer. (Equipment).
New intermeshing mixer vs. traditional mixers.
Optimizing mixing in the Farrel Banbury mixer with wing function.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters