Factory testing and control of raw natural rubber and mixing batches using the RPA.

Technologies

The rubber industry consumes more natural rubber than any single synthetic elastomer. However, the quality of this natural rubber can vary greatly because of its natural origins. Examples of some of the sources of this variation are tree clone; age of tree; soil conditions; climate conditions; manner of latex coagulation; processing history; and storage history.

For example, the Hevea tree clone can produce bimodal or unimodal molecular weight distributions (refs. 1 and 2, figure 1). Also, it has been reported that the magnitude of molecular weight distribution (MWD) can vary greatly with the [M.sub.W]/[M.sub.N] ratio between 2.5 to 10 for all grades of NR (ref. 3). Moreover, the reported gel content for natural rubber (believed to be caused by free radical crosslinking [ref. 4] or reaction of aldehyde groups with free amino acids or protein molecules [ref. 5]) can vary from 5% to 50%. These quality variations are very significant and are a result of the different conditions given above.

[ILLUSTRATION OMITTED]

The purpose of this study was to quantify the variation of raw NR within a factory environment by characterizing the extent of that variation as it relates to viscoelastic properties and then to examine the effects this variation has on the tire manufacturing process.

Experimental

In order to measure the variation of raw natural rubber and to quantify the variations in the mixed compounds due to the rubber, a rubber process analyzer (RPA 2000 Auto) was set up for factory testing. The RPA was applied to testing bales of raw natural rubber, the mixing of the masterbatches, the remills and the final batches for a NR based wire coat stock.

The following sections describe how the samples were taken during two factory mixing campaigns - samples of raw NR polymer and mixed compound. Also, the mixing procedure used by the factory and the RPA test conditions are described.

Factory mixing conditions

This study was performed in cooperation with the factory mixing of a tire wire coat compound. This wire coat formulation was based on 100 phr NR (SIR20). Three mixing stages were used for each batch - a masterbatch, a remill and then a final mix stage. Each mix was performed at constant rotor speed using internal mixers. All batches were dumped once a specified temperature was reached. The time and total energy at dump (integrated power at dump) were automatically measured for every batch.

Sampling procedure

For raw natural mbber sample collection, one sample of the SIR 20 NR was taken at every fifth mix. Three specimens from each sample were tested with the RPA in order to measure within bale variation.

Also, one sample was taken of mixed stock from each completed batch. However, in order to measure within batch uniformity, a total of 15 samples were taken from the last batch of the mixing campaign (run).

Test conditions

The RPA testing was applied to two separate mixing campaigns (or runs) about four days apart. The fast campaign included 63 batches, while the second consisted of 76 batches.

The RPA was configured to perform a test of short duration, just over two minutes, in order to complete each test within the mix cycle time and keep up with production. (The fastest mix cycle recorded for this compound was 2.5 minutes).

The RPA 2000 Auto was used for all factory testing. This automated unit was configured with PET polyester (27 [micro]m thickness) film for the upper layer and 51 [micro]m thick Nylon 6,6 film for the lower layer. (Through this combination of films, an applied strain of 100% can be applied to the samples without tearing the film). Only one RPA test configuration was used throughout this study. It consisted of a short one minute conditioning at 0.5 Hz, 1% strain and 100 [degrees] C, followed by a simple strain sweep at 1 Hz and 100 [degrees] C. This strain sweep consisted of 1, 10, 30 and 100% strain. The total lapse test time is two minutes and seven seconds. In order to keep these tests short, no low frequencies were applied.

The following are the test parameters that were reported here from this RPA test:

* S' elastic torque;

* G' elastic modulus (storage modulus);

* G" viscous modulus (loss modulus);

* tan [Delta] [=(G"/G') = (S"/S')].

The following are other test parameters that were available from the RPA, but not reported here:

* S" viscous torque;

* S* complex torque;

* G* complex dynamic modulus G* = [[[(G').sup.2] + [(G").sup.2]].sup.1/2];

* [Eta]' real dynamic viscosity;

* [Eta]* complex dynamic viscosity.

Discussion of results

Raw natural rubber variability

Between bale variation was simply measured by calculating the standard deviation (S) of the averages of the triplicate testing of each bale.

Within bale variation was calculated by pooling the standard deviations from the triplicate testing of each bale as shown in equation 1.

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Where: [S.sub.P] is the pooled standard deviation; [S.sub.1], [S.sub.2], [S.sub.3], ... [S.sub.n] are calculated standard deviations from triplicate testing of bales 1, 2, 3, ... n.

In both cases, results were expressed as coefficient of variation (CV) which is simply the percent of the mean that is represented by one standard deviation. This is given in equation 2.

(2) CV = 100 [multiplied by] ( std. dev./mean)

The percent range of variation (RV) is calculated from equation 3.

(3) RV = 6 [multiplied by] CV

Statistically, the RV represents the probable range at a 99% confidence level that these data will display such a range.

Magnitude of natural rubber - within bale variation

Table 1 gives the coefficients of variation for within bale variation for the elastic modulus (G') and the viscous modulus (G") at three different strains for both mixing campaigns. As can be seen from this table, the within bale variability from the CV calculations is about 2-4% for G', and 2-3% for G". As a percent range of variation (RV) for within the bale, G' is about 18%, while G" is about 12%.

Table 1

Test                        G'@10%   G'@30%   G'@100%

First run
CV within bale for G' (%)     4.19     4.29      4.56
CV within bale for G" (%)     3.84     3.40      2.73

Second run
CV within bale for G' (%)     2.64     2.82      3.58
CV within bale for G" (%)     1.98     2.41      2.04

Magnitude of natural rubber - between bale variation

Table 2 gives the CV for between bale variation for these same properties. Moreover, the coefficients of variation between bales is somewhat higher at 4-7% for G' and 3-4% for G". This translates into a percent range of variation of about 33% (or as high as 42%) for G' and about 20% for G".

Table 2

Test                     @10%   @30%   @100%

First run
Between bale CV G' (%)   4.60   5.21    7.64
Between bale CV G" (%)   3.77   4.12    3.09

Second run
Between bale CV G' (%)   4.40   4.92    6.06
Between bale CV G" (%)   3.46   3.68    2.02

Magnitude of natural rubber - between campaign variation

Figures 2 and 3 show run charts of the bale data for the first and second mixing campaigns, respectively. These run charts show not only the variation of the SIR 20 natural rubber, but also the variation from a parallel study of bales of SBR 1500 which were used in another compound. This comparison of the variation of natural rubber vs. SBR shows quite a contrast. It confirms that the RPA displays good repeatability and bales of natural rubber are much more variable than the bales of synthetic SBR. Any variations in NR data are therefore due to the polymer, and not to the test procedure, nor to the test instrument.

[GRAPHS OMITTED]

Strain dependency of bale-to-bale variations

Figures 4 and 5 compare the bale-to-bale variations of G' and G" for three different strains (10, 30 and 100%). Figure 4 makes these comparisons for the first mixing campaign and figure 5 is from the second campaign. (The 1% strain condition was also used in this study but was found to be too noisy for viscous measurements.) Even though these data are from two separate mixing campaigns, they show good agreement. It can be seen that with each mixing campaign, greater variations were measured by the elastic modulus G' when the applied strain was increased to 100%. From previous studies it has also been observed that higher discrimination to NR differences was achieved by applying higher strain (ref. 6). An example of this is shown in figure 6. Overall, an increase in discrimination power to NR variation is achieved when a higher strain is applied. Therefore, 100% strain is the preferred test condition for measuring raw natural rubber quality. Also, it may be noted from figures 4 and 5 that there is significantly more NR variability seen from the RPA elastic response than from the viscous response when measured at higher strains such as 100%. This has also been observed in previous ASTM studies (ref. 7). Figure 7 illustrates this difference. As can be seen, typically different grades of natural rubber will differ more in elastic quality than viscous quality. This is probably because of significant variations in gel content or chain entanglements. Either of these quality characteristics will greatly affect the elastic response from viscoelastic measurements. The high strain capability of the RPA allows these significant differences to be quantified, allowing it to be used as a sensitive quality tool to assess NR variability. Other instruments at lower strains will appear to measure uniformity, and not the variabilities that exist, and are detectable at high strains.

[GRAPHS OMITTED]

NR compound variability

One sample each was taken from every batch mixed for the masterbatches, remills and finals of the wire coat stock for each of the two mixing campaigns (runs). This was done to measure batch to batch variation. Also, in order to measure within batch uniformity, 15 samples were taken from the last mixed batch of the mixing campaign. About 500 specimens were measured in total.

Batch-to-batch uniformity was simply calculated from the standard deviation of the results from the individual RPA measurements for each batch mixed, while within batch uniformity was calculated from the standard deviation for the 15 samples from the last batch mixed in the mixing campaign.

Figure 8 shows a control chart of the first masterbatch mixing campaign consisting of 63 consecutive batches. Also shown is the variation from testing the last batch (number 63) a total of 15 times. This shows graphically that the batch-to-batch variation is significantly high; however, the within batch variation is much smaller.

[GRAPH OMITTED]

Table 3 quantifies this batch-to-batch variation and the within batch variation for the first and second mixing campaign. Figure 9 graphically shows very good within batch uniformity (CV [is less than] 1%), but between batch variability is still significant with a CV of 3% (equal to a % range of variation of 18%), even though it is less than the variability measured for between bales of raw natural rubber. Also, this figure shows that the large variability of the NR polymer has been reduced somewhat by the mastication process, and that the compound shows about the same level of variation even when tested at higher strains.

[GRAPH OMITTED]

Table 3

First mixing campaign                 @10%   @30%   @100%

CV of G' for within batch variation   0.79   0.54    0.89
CV of G' for between batch            3.33   3.42    3.17
  variation
CV of G" for within batch variation   0.94   0.78    0.43
CV of G" for between batch            1.84   2.07    1.9
  variation

Second mixing campaign                @10%   @30%   @100%

CV of G' for within batch variation   3.47   3.26    2.54
CV of G" for between batch            2.37   2.37    2.13
  variation

Figure 10 shows the change in the average values for G', G" and tan [Delta] for the raw natural rubber, the first pass of the wire coat compound in the mixer, the remill and the final mix. The G' and G" both are higher for the wire coat compound than for the raw NR because of the presence of carbon black and other compounding ingredients. As can be seen with increasing mixing work history, the G' and G" both decrease; however, the tan [Delta] increases with a higher state of mix (more work history). Figure 11 shows the reduction in bale-to-bale and batch-to-batch variability with increasing work history going from the raw polymer to the final mix. As shown, there is a decelerating improvement in uniformity with increasing work history.

[GRAPHS OMITTED]

Time to dump and total energy at dump

All batches were dumped at a predetermined temperature. Figure 12 shows the total energy at dump (the integrated power over the entire mixing cycle) and mixing time to dump are directionally related. There are reasons for this positive correlation. It is likely that energy at dump is a function of the complex modulus G*. The rotor was turned at a constant speed. The power measured the input needed to cause the rubber to flow (governed by the viscous modulus), and the input needed to overcome the interactive forces of the rubber (governed by the elastic modulus). As discussed earlier, G* is the square root of the sum of the squares of G' and G", and sums both the viscous and elastic moduli. Tougher rubber (higher G*) takes more energy to mix, but may not have such a high viscous quality to raise the mix temperature as quickly. Therefore, these two mixing characteristics moved together in our study.

[GRAPH OMITTED]

As seen in figure 13, G" @ 100% strain for the mixed stock inversely correlates better to the total mixing energy at dump with an inverse correlation (R = -0.80) than G' does at 100% applied strain (R = -0.66). If the resulting viscous modulus G" of a batch after mix was higher, it suggests that applied energy to the stock during mixing was more rapidly turned into heat, therefore causing the batch to reach the dump temperature more quickly. This could explain the inverse correlation between G" @ 100% strain and mixing time, as well as total energy at dump. Also, the higher strain at 100% for testing these mixed batches is needed here to relate well to the energy at dump. RPA tests for G" at 30% strain gave a poorer correlation (R = -0.67) while RPA testing for G" at 10% and 1% strain gave no correlation (R [approximately equals] 0).

[GRAPH OMITTED]

Raw natural rubber properties vs. total energy at dump

Figure 14 shows that G' @ 100% strain for bales of raw natural rubber relates very well to the total energy at dump for the batches in a mixing campaign. The higher the G' @ 100% of the raw NR, the lower the resulting total energy at dump from mixing. This is probably due to a higher initial shearing and faster breakdown during the mixing process. This has been quantified in an earlier study (ref. 8) in which the rate of decrease in elastic torque with mastication was measured vs. the initial Wallace Plasticity (Po). This same study observed that the Po correlated better to the elastic response rather than the viscous response. It was observed that the higher the initial Wallace Po and the higher the elastic modulus G' responses, the faster the decrease in rubber elasticity due to mastication. This may explain the inverse relation observed in figure 14.

[GRAPH OMITTED]

Table 4 shows that a higher strain needs to be applied to raw natural rubber in order to get a better correlation with total energy at dump during mixing. Also, it is important to observe that while the elastic modulus G' is inversely related to the total energy at dump and the total mixing time, the viscous dynamic modulus is positively correlated to energy at dump and mixing time. A higher viscous quality means that more energy is absorbed during the mixing process and more power is required to keep the mixing rotors turning at the same velocity.

Table 4

Correlation coefficients of total energy @ dump vs. G' at
different applied strains for the raw natural rubber

Applied strain   (-) R

      1%          0.17
     10%          0.82
     30%          0.82
    100%          0.85

From these relationships, the factory can take measurements from the raw bales of natural rubber in advance of the next batch to be mixed. Mixing conditions could be adjusted beforehand in order to obtain more uniform and consistent mixing.

Conclusions

* The rapid two minute test programmed on the RPA Auto was very effective and practical for testing the quality of both raw natural rubber bales and factory mixed stocks.

* The within bale coefficient of variation (CV) for the elastic quality for the SIR 20 NR used in this study was 2-4% with a % range of variation (RV) of 12-24% compared to the CV for within bale variation for SBR 1500 which was measured at 1%.

* The within bale CV for the viscous quality for the SIR 20 used was 2-3% with a calculated RV of 12-18%.

* The between bale CV for the elastic quality of the SIR 20 NR used in this study was a CV of 4-7% with a calculated RV of 24-42%.

* The between bale CV for the viscous quality of the SIR 20 NR used in this study was 3-4 % with a calculated RV of 18-24%.

* Natural rubber normally varies more in its elastic quality than its viscous quality.

* The natural rubber variation contributes to the viscoelastic variability of the compounded batches, but additional batch work history through milling and additional internal mixer passes reduces this variability to less than half.

* RPA testing at the higher strain of 100% is more sensitive to real viscoelastic variations for both the raw natural rubber, as well as the mixed batches.

* The elastic quality (G') at high strain (100%) for different raw natural rubber bales appears to be a better predictor of the resulting mix cycle and mix quality than the viscous quality of raw NR.

* The total energy at dump for the mixing process appears to relate better to the resulting viscous quality (G") for mixed individual batches than the elastic quality G' in this study.

References

(1.) G.M. Bristow and B. Westall, "Molecular weight and molecular weight distribution of natural rubber," Polymer, London, 8, 609.

(2.) A. Subramaniam (1976), Gel Permeation Chromatography of Natural Rubber, Rubber Chem. & Technol., 45, p. 345-358.

(3.) RRIM Technology Bulletin 4, Molecular weight and molecular weight distribution of natural rubber.

(4.) Ibid, p. 105.

(5.) A. Subramaniam, "Mooney viscosity of raw natural rubber for greater consistency," Rubber World, July, 1988, p. 271.

(6.) "Rubber characterization by applied strain variations using the rubber process analyzer," by J.S. Dick and H.A. Pawlowski, Rubber World, Jan. 1995.

(7.) "Quality assurance of natural rubber using the rubber process analyzer," by J.S. Dick, C. Harmon and A. Vare, ACS, Rubber Div., May, 1997 (paper no. 97), J. of Polymer Testing, Sept., 1999.

(8.) Ibid, Figs. 27-28.

Christopher A. Stevens and John S. Dick, Alpha Technologies