Printer Friendly

Utilization of the rubber process analyzer in Six Sigma programs.

The Six Sigma process was originally implemented by corporations such as Motorola and General Electric (refs. 1 and 2). It has evolved into a rigorous industrial problem solving process, using a very practical application of statistical methodologies such as well planned design of experiments (DOE) and other statistical analysis techniques to improve factory quality, increase product output and reduce manufacturing costs. If a manufacturer can run their rubber process at a "process sigma" of 6, this means that their DPMO (defects per million opportunities) are only 3.4 (these calculations are made with actual process data and customer specifications and then scaled up to the equivalent of a million opportunities). This process sigma is used as the measure of quality where the Six Sigma philosophy is such that "even one detect is a failure in your customer's eyes" (ref. 3). By contrast, traditional processes many times have DPMO exceeding 60,000. If a typical process is running a scrap rate of 1%, this translates into 10,000 parts per million.

A sensitive gauge

One of the key requirements of the Six Sigma process is to have an effective, sensitive gauge. Many Six Sigma projects have been launched in the rubber industry in order to find assignable causes of variation and take corrective action to improve the efficiency of the process with a reduction in scrap and rework. When these methods are applied to problems involving metallic materials used to make a rubber product, the Six Sigma teams usually are quite successful because the traditional test methods are very sensitive to process variation. However, when a Six Sigma team investigates a problem relating to rubber variation itself, many times they have a harder time determining assignable causes and taking the necessary corrective actions to solve the problem. Experience has shown that this difficulty is usually due to the utilization of an insensitive rubber test method. However, much success has resulted when the RPA rubber process analyzer (hereafter referenced as the RPA) is used in this process. Figure 1 illustrates this point. As can be seen from this figure, when traditional test methods such as maximum ([M.sub.H]) and minimum ([M.sub.L]) torque from the MDR, and ultimate tensile strength and tensile modulus testing are applied to a rubber process, they could not effectively separate two quite different factory mixing campaigns of the same compound. On the other hand, as this same figure shows, the RPA is very effective at showing the significant differences between the first and second mixing campaigns as two well defined, separate populations. The other traditional test methods are not sensitive enough to separate these batches into two separate populations.


Many times the Six Sigma process requires identification of assignable causes of variation for a robber mixing process. Sometimes these causes of variation are found in the raw rubber. From an earlier Rubber Manufacturers' Association (RMA) study, two nitrile polymers were found to be the same by all traditional rubber tests including Mooney viscosity and percent ACN content. However, these two nitrile polymers repeatedly processed differently in the factory. Testing these two polymers with a new, rapid RPA test configuration showed that their viscoelastic properties were greatly different. The RPA displayed superior statistical test sensitivity as measured from signal-to-noise (S/N) calculations (ref. 4). In this case, for a test method to be discerning and effective in identifying the "good" vs. the "bad," one should have a signal to noise ratio above six. Only the RPA was effective in producing S/N ratios well above six. Other traditional test methods did not show this sensitivity and are therefore ineffective for quality assurance and Six Sigma programs for these types of nitrile polymers.

RPA and Six Sigma programs applied to the rubber process

As is well known, the rubber manufacturing process is composed of several specific process stages. These process stages may consist of any combination of the following: Receiving of raw polymers and compounding ingredients, storage; first pass mix; second pass mix: calendering; cold feed or hot feed extrusion; compression, transfer or injection molding; and continuous vulcanization or autoclave curing, etc.

Typically, a detailed process mapping of the process stages is made by the Six Sigma team in order to have a thorough understanding of how the process works. No two processes are exactly the same. The Six Sigma team that performs this process mapping should be from preselected areas of manufacturing, quality assurance, R&D, accounting, as well as other areas as needed (just as Motorola initiated teams in the early 1990s).

Examples of rubber product problems which may prompt a full Six Sigma program with the RPA are as follows: Blisters; bare spots; porosity; delamination; off spec. dynamic properties; off spec. tensile properties; off spec. hardness; non-fills; under-fills; out-of-spec, cured shrinkage; poor dimensional stability; die swell variation; pin holes; lumps; undispersed agglomerates; and variable cure profiles

These problems which may occur either downstream or midstream in the manufacturing process, usually have assignable causes of variation further upstream. By applying the RPA testing to establish a baseline for viscoelastic properties at different stages of the rubber manufacturing process, the data can be used to find assignable causes of variation for downstream problems, so that the appropriate corrective actions can be taken.

It is usually important that the Six Sigma team decide beforehand exactly where in the processes to take samples for RPA testing (sampling points) and exactly what RPA test configurations should be implemented. The Six Sigma team should consist of factory associates from key relevant areas of the plant operation. Clear objectives should be established, and a timeline and a Gantt Chart should be established for completing the Six Sigma project.

Measurement of raw rubber variation

The traditional method for measuring the quality of raw rubber is to perform a Mooney viscosity test. However, it is now well known that the Mooney only partially appraises the quality of the raw polymer. Figure 2 gives an example where two sources of the same SBR have the same Mooney viscosity value but show different viscoelastic profiles from RPA analysis which predict different processing characteristics. These types of comparisons have also been reported for other raw elastomers as well as natural robber or NBR (refs. 5-8). These viscoelastic differences in raw polymers, which the Mooney viscometer can not detect, can have a large effect on the processability of rubber compounds downstream.


Raw NBR variation

Figure 3 shows the viscoelastic differences which were quantified from a RMA RPA interlaboratory crosscheck study of two different lots of the NBR (105 vs. 106) from the same manufacturer with the same Mooney viscosity and the same ACN content. A further RMA study quantified the following compound performance differences when they were mixed in formulations. The poor processing NBR (105) was reported to behave as follows in relation to the good processing NBR (106) (ref. 9):


* 66% longer black incorporation time,

* 20% higher die swell:

* higher mill shrinkage:

* poorer extrusion surface appearance.

Raw natural rubber variation

Figure 4 shows two natural rubber SIR 10 natural rubber sources with the same Mooney viscosity. These two samples were submitted as part of an ASTM study on definitive testing of natural robber (ref. 10). Even though these two NR samples possessed identical Mooney viscosity, they processed quite differently in the factory internal mixer. RPA testing shown in this figure indicates viscoelastic differences which can be seen under high strain test conditions. In this case, variations in elasticity greatly affected mixing characteristics. Previous work indicates that normal variation from lot to lot is significantly greater for this uncured elastic quality than for the viscous quality (ref. 11).


Figure 5 demonstrates quantitatively how uncured elastic variation in natural rubber bales can impact the mixing process for a factory tire wire coat stock being discharged from a mixer at a constant dump temperature. So, the "tougher" the natural rubber results in a lower total energy at dump and a shorter factory mix cycle (ref. 12). This means that the tougher or more elastic natural rubber breaks down more quickly during the factory mixing process. This is demonstrated from another ASTM study of different grades of natural rubber in which it was discovered that natural rubber grades with higher elasticity broke down faster in a laboratory internal mixer study (ref. 13). What is also interesting is that the uncured viscous (or loss) modulus G" of the raw natural rubber bales used in the factory mixing campaign tends to have the opposite effect.


Raw natural rubber shipments can vary considerably in their viscoelastic characteristics. Figure 6 shows the typical variability that can be expected for SIR 20 bales being used in production. Here we can expect a statistical range of 45% (based on 6 x coef. of variation) which has a definite effect on the uniformity of the mixing process. Some rubber fabricators arc beginning to use rapid RPA configurations to test and sort incoming natural robber skids. These sorted bales are dispositioned to different applications (compounds) in the manufacturing process.


Mixing and processing compounds

Often the variation we have noted for raw robber can and does contribute to batch to batch variation in a mixing campaign as noted in figure 7. This figure shows the variation which exists among the bales of raw natural robber, along with the internally mixed masterbatches, the remills and the final batches. As can be seen, the high variation of the raw natural rubber greatly contributes to the variation of the masterbatches from the first pass of this wire coat stock. However, this variation is reduced somewhat with the mixing of the remills and then the final batches through effective mixing techniques. As more and more work history is applied to the rubber, a greater reduction in the uncured elasticity also occurs which contributes to better downstream processing. With natural rubber mixing, the uncured elasticity generally drops more rapidly than the viscous quality of the rubber compound. This usually means that the uncured tan [delta] (G"/G') will increase from greater work history.


On the other hand, many times batch to batch variation is not caused by raw rubber variation at all. This is demonstrated in figure 8 with the mixing of a synthetic factory tire tread stock. As noted, there is very little bale-to-bale variation observed for the SBR 1712 used as the main rubber in the mixing of this tread stock. However, the variation from the completion of the masterbatch is significant. On the other hand, the variation from batch-to-batch tot the finals is somewhat less than observed for the masterbatches because of the extra work history. Fine particle size carbon blacks and silicas often require additional work history in order to achieve good dispersion and good batch to batch uniformity. In fact, consistent incorporation and dispersion of fine particle sized fillers can be a commonly occurring problem with many Six Sigma programs. This has been reported for factory mixing campaigns of multiple batches of a silicone rubber compound using a doughmixer (ref. 14). This batch to batch variation can be very high. This is generally the case because of the problem in achieving good consistency from batch to batch due to the difficulty in achieving good dispersion of fumed or precipitated silicas.


So far we have shown process variation with a statistical range from 6 to 70% variation. What is acceptable? That question can only be answered by the sensitivity of the downstream processes such as extrusion, calendering, injection molding, compression molding, continuous vulcanization, etc. It is well known that excessive variation in the uncured elasticity of mixed batches can and does lead to downstream quality problems such as dimensional stability problems with extrusion, non-fills in injection molding and physical property variations with continuous vulcanization, to name a few. Measurable internal or external failure costs (scrap, rework and/or returned goods) are many times caused by excessive variability in the output from the rubber mixing process. Typically, in a Six Sigma program with the RPA, a baseline for this variability is established using the RPA measurement capabilities just illustrated. Usually, this baseline is established using the principles of statistical process control (SPC).

Baseline and SPC

Once this baseline is established, as well as a con-elation with downstream quality problems, a Six Sigma team (representing different areas of manufacturing, compounding and laboratory testing) brainstorms on possible assignable causes of this variation. Usually, a relationship can be established between the RPA processability parameters (from such test configurations as described in ASTM D6204) and mixing parameters from factory data acquisition systems. Also, these teams will usually discuss how this variation might be reduced, as well as what the appropriate targets and specification limits should be for key RPA test parameters.

Finding the upstream resolution to the downstream problem

Part of the challenge of the Six Sigma process is for the team to find the upstream root cause for a downstream problem so that a permanent solution can be found. Many of the statistical tools taught in the Six Sigma process are used to find the solution to the downstream problem. This process will usually involve extensive RPA factory data collection. Sometimes limited design of experiments (DOEs) will be conducted. Many times more than one DOE will be required to find the ultimate solution. It is quite common for a team to work on a factory problem for several weeks before it is finally corrected.

Work history

Each stage in rubber fabrication usually adds additional work history to a rubber compound before that robber ,arrives at the cure process. This is true whether that stage is a first pass, second pass or third pass internal mix, milling, extrusion or calendering. Work history can be determined by integrating a power-time curve from a mixing, extruding or milling process. Usually, a power integrator is used to determine a total applied work history for a process. For most formulations, additional mechanical work applied to the com pound results in the uncured elasticity decreasing. Reducing this elasticity makes the rubber stock less nervy and usually results in better processing further downstream. This decrease in elasticity is illustrated in figure 9 with a drop in a SBR/BR tire tread's elasticity from increased work history.


Usually, in a similar manner, the viscous quality of the rubber compound will also decrease with additional work history, but many times this quality does not drop as quickly as the elastic quality just described. As a result, the uncured tan [delta] (= viscous modulus G"/elastic modulus G') will usually (but not always) rise with increased work history. In fact, there are a few cases where tan [delta] can be even more sensitive to changes in work history than G'; however, many times the opposite is true and G' is more sensitive. This is very compound dependent.

Traditionally, rubber compounds have been dumped from mixing equipment by time and temperature. However, a better way to monitor the state-of-mix in an internal mixer is to use a power integrator. Total energy at dump recorded by a power integrator is a more effective method of controlling the robber mixing process than traditional time and temperature techniques. This assumes the addition sequence and time-temperature profiles remain relatively unchanged (ref. 15). Therefore, monitoring the total energy at dump (kWh) is a very effective method of recording the total mechanical work history that is applied to the rubber compound during the mixing process. Also, the RPA test parameters can be used to monitor the effects of additional applied work history on the rubber compound's viscoelastic properties and processability characteristics. This can be illustrated from earlier experimental laboratory work (ref. 16) with an all natural rubber formulation, as shown in figure 10. These figures demonstrate how the elasticity decreases very quickly with additional work history, while the viscosity also decreases, but at a slower rate. Since the uncured tan [delta] is a dimensionless ratio of the viscous quality (G") divided by the elastic quality (G'), the denominator is decreasing faster than the numerator. So, the tan [delta] in this case actually rises with additional work history. Sometimes, however, the elasticity does not decrease faster than the viscous quality with some other compounds. In those cases, one should study and monitor the changes in elasticity with time.


Because of the cited changes in viscoelastic quality of a rubber batch with increasing applied work history, the uncured G' elastic modulus has become a good indication of the relative state-of-mix (a lower G' usually implies a better quality of mix). In mixing a final, a higher state-of-mix also implies many times a better state-of-cure, as illustrated in figure 11. This is simply because a better state of mix means that the curatives are dispersed better and a better state-of-cure is achieved through a higher crosslink density (ref. 17).


Heat history

Monitoring and controlling heat history of a compound in the plant can be as important as controlling total work history. Just as with work history, the RPA can be used to monitor heat history applied to a compound by measuring the changes in compound scorch safety and cure times as the final mixed batches pass through downstream processes. Significantly greater test sensitivity to variations in scorch safety has been achieved by using the variable temperature analysis (VTA) feature of the RPA. By programming the RPA to linearly ramp the test temperature from a typical process temperature of 100[degrees]C to a cure temperature of 195[degrees]C in 10 minutes, a 55% improvement in statistical test sensitivity was measured with silicone compounds (ref. 18).

As discussed previously, heat history can be determined from measuring the time-temperature profile for a given rubber stock at different stages of factory processing. The effects of the measured time-temperature profile on the scorch and cure time characteristics of a factory rubber stock can be determined through the RPA either empirically or through the use of cure kinetics.

Empirical approach

If a time-temperature heat history profile tot a rubber compound during a robber process(es) can be established, then the RPA variable temperature analysis can be programmed to simulate this profile to study its effects on compound scorch and cure characteristics (ref. 19). This is a good empirical method for determining the effects of variations in heat history on the scorch and cure characteristics of a robber compound. Not only can a time-temperature profile from a factory cure be simulated by RPA VTA, but also profiles including milling, extrusion or calendering can be combined with curing to make better predictions. Also, variations in heat history associated with injection molding can also be entered into the VTA of the RPA for more accurate predictions on scorch and cure (ref. 20).

Cure kinetics approach

This is another method which can be used to predict the effects of variations in heat history from processing the rubber at different stages on the final scorch and cure properties. To start off, the RPA is a very effective rotorless curemeter with a short temperature recovery time. This makes it an effective instrument for measuring the change in reaction rate constants when a series of separate cure tests are run on the same compound at different temperatures. By knowing the order of reaction and measuring the slope of these different reaction rate constants vs. 1/[degrees]K, one can calculate the compound's specific energy of activation ([E.sub.A]) through what is called an Arrhenius plot (refs. 21 and 22). Once the [E.sub.A] is determined, the time-temperature profile can be partitioned into segments and cure equivalents can be calculated for each segment using the Arrhenius equation. By summing the cure equivalents over time, the heat history of the factory compound can be modeled.

Work history and heat history interdependence

As just discussed, work history and heat history are both very important; however, they are also very interdependent. For example, if you apply a great deal of work history to a rubber compound in the factory, the temperature of this rubber compound will increase. This is called viscous heating from mechanical friction. Different rubber compounds have different propensities to viscous heating from a given level of mechanical work input. The RPA can be used to directly measure a compound's propensity to have viscous heating from mechanical deformation (ref. 23). Figure 12 shows RPA viscous heating measurements for rubber compounds containing different types of carbon black and other fillers. The RPA can directly measure the viscous heating of a rubber compound with good repeatability (ref. 24).


Also, changes in rubber stock temperature affect the level of work that can be applied. It is well known that increasing the temperature of a rubber stock will also reduce its viscosity. This effect is demonstrated in figure 13, which shows the drop in viscosity of various compounds based on a wide variety of different base raw elastomers when the temperature is increased. An old rule of thumb is that a 10[degrees]C rise in temperature will reduce the viscosity by approximately 10%. As can be seen, the validity of this role of thumb very much depends on the base elastomer. Sometimes this rule may be true while other times it is not. The important point to note is that the viscosity always decreases with a rise in temperature. This phenomenon is very important when rubber is being mixed in an internal mixer. At the start of the mixing cycle, the robber is cool and somewhat resistant to flow (and mixing). This means the motor draws more power (energy) to allow the rotors to turn at the pre-set rpm speed. With the rotation of the rotors, internal friction of the rubber causes viscous heating and raises the batch temperature. As the batch temperature rises, its viscosity decreases and the amount of energy required to turn the internal mixer rotors at the pre-set speed is reduced. There fore, as the temperature of the rubber batch increases, the rate of work history input will decrease. So the rate of viscous heating is effecting the rate of applied work history. In fact, during the later stages of the mix, the viscosity will drop so much because of the higher temperature that not enough shear is being provided for effective mixing. At this point, the batch is usually dumped. So work history and heat history of a batch during mixing are very interdependent.


This interdependence of work history and heat history can be seen clearly from a long mixing campaign (mixing multiple batches of the same compound), it is not uncommon to see this relationship in what is called the first batch effect. If one begins a mixing campaign with a relatively cold mixer and consistently dumps by the same preset dump temperature, the first batch will take significantly longer to reach the dump temperature and will therefore receive more work history. This is illustrated in figure 14. As the mixer beats up, the time required for the following batches to reach the preset dump temperature may become less and also the total work history becomes less. Also, as discussed earlier, less work history is measured as higher uncured elastic modulus G'.


It should also be noted that many factory internal mixers have very efficient cooling systems and therefore may not show this so-called first batch effect.

Six Sigma solutions

Once a correlation is identified between downstream quality problems and upstream viscoelastic properties measured with the RPA, the Six Sigma team will work on finding the root cause(s) and what are the most cost-effective long-term corrective actions to be made. Many times, brainstorming techniques are used with cause and effect diagrams and Paredo charts. After execution of design of experiments (DOEs) or more limited factory trials, proposed changes or corrective actions can be verified or disproved. If the proposed corrective actions do not work, others are tried until scrap levels are reduced or eliminated. Usually these efforts are ongoing.

Over 15 years ago, some companies were using a traditional quality philosophy that there was a law of diminishing returns in quality efforts based on traditional quality costs (ref. 25). The Six Sigma philosophy (ref. 26) states that continued effort at finding root causes of quality variations will actually result in far greater quality cost savings in the long run.


The following are the conclusions from this study:

* Factory experiences have proven that the RPA is a very effective gauge for measuring rubber process characteristics for a Six Sigma program.

* The RPA was found to be significantly more sensitive to raw robber and mixed batch quality variations than traditional test methods such as Mooney viscosity, capillary rheometer, MDR, density and ultimate tensile strength and modulus.

* The RPA has proven to be an effective measure of both work history and heat history applied to a robber process.

This article is based on a paper given at the October; 2001 meeting of the Rubber Division.


(1.) F.W. Breyfogle III, Implementing Six Sigma, Smarter Solutions Using Statistical Methods, John Wiley & Sons, Inc., NY 1999.

(2.) Thomas Pyzdek, The Six Sigma Handbook, Complete Guide, McGraw-Hill, NY, 2001.

(3.) Rath and Strong Management Consultants, Six Sigma Pocket Guide, 2000. p. 89

(4.) W. Cousins (Bayer), and J. Dick, "Effective processability measurements of acrylonitrile butadiene rubber using rubber princess analyzer tests and Moaner stress relaxation," Rubber World, January, 1998.

(5.) J.S. Dick and H. Pawlowski, "Applications for the rubber process analyzer," Rubber and Plastics News, April 26 and May 10, 1993.

(6.) W. Cousins and J. Dick "Effective processability measurements of acrylonitrile butadiene rubber using rubber process analyzer tests and Mooney stress relaxation," Rubber World, January, 1998.

(7.) J.S. Dick and H.A. Pawlowski, "Rubber characterization by applied strain variations using the Rubber Process Analyzer," Rubber World, January, 1995

(8.) J. Dick, C. Harmon and A. Vare "Quality assurance of natural rubber using the rubber process analyzer; Polymer Testing, 18 (1999) 327-362.

(9.) W. Cousins and J. Dick "Effective processability measure meats of acrylonitrile butadiene rubber using rubber process analyzer tests and Mooney stress relaxation," Rubber World, January, 1998.

(10.) J.S. Dick, "Progress report on natural rubber testing performed for the task group on NR definitive testing," ASTM D11.22 Task Group Report, November 1994.

(11.) J. Dick, C. Harmon and A. Vare "Quality assurance of natural rubber using the rubber process analyzer, Polymer Testing, 18 (1999) 327-362.

(12.) C. Stevens and J. Dick, "Factory testing and control of raw natural rubber and mixed batches using the rubber process analyzer," Rubber World, January,, 2001.

(13.) J. Dick, C. Harmon and A, Vare, Op Cit

(14.) John S. Dick, Chris A. Sumpter and Brian Ward, "New effective methods for measuring processing and dynamic property performance of silicone compounds," KGK (Kautschuk Gummi Kunststoffe), September, 1999 (52 600-607).

(15.) "Power integrator; a more precise and efficient method for control of batch-to-batch rubber processing and property uniformity," Monsanto Technical Bulletin, IE3 p. 1.

(16.) J. Dick and H. Pawlowski, "Applications for the rubber process analyzer, parts 1 and 2," Rubber and Plastics News, April 26 and May 10, 1993.

(17.) J. Dick, "The optimal measurement and use of dynamic properties from the moving die rheometer for rubber compound analysis," Rubber World, January, 1994 and Revista Caucho, February, 1996.

(18.) John Dick, Chris A . Sumpter and Brian Ward, "New effective methods for measuring processing and dynamic property performance of silicone compounds," paper No. 10, presented at the ACS Rubber Drip, Sept. 1998.

(19.) J. Dick and H. Pawlowski, "Application of the rubber process analyzer in characterizing the effects of silica an uncured and cured compound properties (with H.A. Pawlowski)," ITEC '96.

(20.) John Sezna, "'RPA testing for injection molding of rubber," paper no. 173, Rubber Div. ACS, Sept., 1999.

(21.) John Sezna and W. Curtis Woods, "Thick article cure prediction," presented at the Rubber Div., October; 1990

(22.) John Dick and Henry Pawlowski, "Application for the curemeter maximum cure rate in rubber compound development, process control and cure kinetic studies," Polymer Testing 15, (1996) 207-243.

(23.) John Dick, Henry Pawlowski and John Moore, "Viscous heating and reinforcement effects of different fliers using tire rubber princess analyzer," Rubber World, January, 2000.

(24.) Op Cit., J. Dick, Pawlowski and Moore.

(25.) Danuel M. Lundvall, Section 5, Quality Costs, Quality Control Handbook, J. M. Juran (editor). Third Edition, 1974, McGraw-Hill, NY, pp.5-12.

(26.) Thomas Pyzdek, The Six Sigma Handbook, McGraw-Hill, NY 2001, p.166.
COPYRIGHT 2003 Lippincott & Peto, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2003, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Sezna, John
Publication:Rubber World
Date:Jan 1, 2003
Previous Article:Characterization of binary/tertiary blends of SBR, NBR and PVC by IR spectroscopy.
Next Article:Failure analysis of HNBR oil seals.

Related Articles
Burke to lead Goodyear's Six Sigma. (People in the News).
The art and science of winning physician support for Six Sigma change.
Using six sigma principles: using principles developed to eliminate defects in the manufacturing process, the behavioral healthcare industry can...
Characterization of hard-to-process oil field application materials.
Extrusion technology examined.
Six sigma.
Six Sigma for the DoD.
Six Sigma: how it works in Motorola and what it can do for the U.S. Army: how can the Army increase efficiency across its command structure? Put Six...
More business offerings at fall meeting.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters