Custom mixing in the 90s.

Custom mixing in the 90s

Hoover-Hanes, Tallapoosa, GA, is a major custom mixer and compounder for the rubber industry. They supply hundreds of elastomer compounds for a wide variety of end uses. In 1989 Hoover-Hanes corporate management decided to dramatically increase the quality of the products they supply.

This goal was not a reaction to current customer demand. They were responding proactively to the fast approaching manufacturing realities of the 1990s. The management team foresaw these realities as reduced production costs, increased quality of the product they supply and greater customer satisfaction. This was part of their larger goal of becoming a world leader in rubber compounding.

Towards this goal they knew they must have three key ingredients.

* Plant automation

* Plant modernization 

* A corporate philosophy grounded in statistical process control.

Early on, they realized that their first step toward implementing these goals was to accurately collect rheology data and do SPC on that data. To that end, they reviewed their current system, and installed new hardware and software to help them get this under control.

Old method

The previous rheology system was based on flat-bed chart recorders. This type of recorder only plots torque curves. Most companies print a maximum of ten curves per chart as the chart becomes less legible with more than ten. With as many as 70 batches per load, they often had to print seven flat-bed charts for one customer order. This method presented several problems.

* No SPC

* Load reports were very time consuming to generate

* No sheet-to-sheet data continuity

* Any test fluctuation destroyed a whole sheet of information.

* Gate location and spread was a guessing game

* Thousands of paper charts

* No way to analyze all of this information on paper

When the corporate commitments were made, one thing was certain. There was no way they could effectively implement SPC by hand. They barely had time to handle the current paperwork. If SPC was going to happen, an automatic method was mandatory. The top managers knew the benefits of SPC, and were realistic in its implementation. A computerized system was the answer.

Often the lab received customer requests for a report dealing with more than just curve location and shape. They also wanted to periodically gather additional information on a run of material. If the lab needed anything more than just basic curve information they had to get it all by hand. Because of the major time and effort involved, this kind of specialized reporting was very difficult.

Because each flat-bed chart shows curves on only ten batches, it is very difficult to see trends occuring at the end of one sheet and continue to the next sheet. In addition, a test fluctuation ruined the current flat-bed chart. Now, instead of just re-running the current aborted batch test, the lab tech had to rerun all of the batches on that sheet. Since Hoover-Hanes uses six minute tests, they lost up to 60 minutes of time.

Gates under the old system were difficult to place accurately for each compound. The lab people made overlays to place on each flat-bed chart to see if all ten curves were inside of gates. Becuase overlays took a lot of time to develop, they used one overlay for similar compounds. It was simply not practical to redo overlays for each individual compound and customer.

A related problem was that they were drowning in a sea of paper. Hoover-Hanes regularly filled filing cabinets with thousands of flat-bed charts. To assemble and analyze all of this information was simply too large a task. At best, they only knew if the day's current set of compounds were in gate limits. Within this type of system, there was no time to perform process control, or search through past runs for process improvement clues.

Not only had Hoover-Hanes made a corporate commitment to SPC, they knew it was only a matter of time before their customers demanded it. They wanted to show their customers they were forward thinking enough to give them what they needed, even before they ask for it.

The company was searching for a turnkey system that had the flexibility to analyze rheology information in several ways. They did not, by any means, consider themselves computer experts, so they needed one company to come in and install both instruments and software. They were aware of companies who purchased instruments from one company and software from a second company. But there were too many horror stories about the lab being down while the software people blamed the instruments people and vice versa. They wanted one company who could supply both pieces.

For various reasons, software often is an afterthought for most companies when they are purchasing integrated hardware-software systems. In most cases the number one reason is that software is the less expensive of the two. This is often true in the beginning. However, over the long haul software can be more costly when you figure in the costs of installing, customizing, learning and training. To minimize these post-purchase expenses, Hoover-Hanes was looking for a software system that included both items.

The software side of the system had to strike a balance between user-friendliness and flexibility. It had to be easy enough for the operators to use and understand. Because, as in most manufacturing plants, the operators had little previous computer experience, but would enter most of the data. It needed to allow supervisors to examine rheology information by process characteristics (compound, shift, mill number, raw material lot, and so on). It also had to provide for sorting through the data in may different ways. The latter was critical to accomplish the third corporate goal related to SPC.

Supervisors needed to know if the current process is running in a state of control. They also needed to examine historical information for clues as to why certain compounds ran poorly (or very well) in the past. With this information they can fine-tune their compounding process to the highest possible quality levels. The lab manager had other prerequisites for system software as well. They run thousands of batches of material each week, so the software system must efficiently handle this load in memory. They also need to quickly select information on a particular compound that may not have run for two years of more.

New method - Rheologic

After a careful search of the available instrument-software systems, the management team decided to purchase the Monsanto line of products. They had some older model Monsanto instruments and were very pleased with the durability and accuracy of the equipment. To begin with, they bought an ODR-2000 rheometer along with RheoLogic, Monsanto's rheology software system. This combination of instruments and software met their corporate goals and were the `top shelf' products in each category. They felt comfortable purchasing from an established rubber equipment and services company. A Monsanto technical/service representative came to Tallapoosa and installed the new rheometer, a PC computer, and the RheoLogic software. The representative demonstrated how the data feeds directly from the instrument of the software and displays on the screen - as the test occurs. He showed how a lab tech could watch the curve, or perform real time SPC during a batch test.

After instrument, computer and software installation and customization there was an extended training session. It's not enough to just dump a bunch of diskettes on the lab manager's desk. Installation and training must be an important part of the turn-key package. The tech/service representative sat down with the key lab people to show them how to run the system. Within half a day they could run a batch test, display the torque curve on the screen as it compared to customer gates, and print a report.

After gathering information on the available hardware/software systems in the market, Hoover-Hanes found that these systems differed in specific ways.

Ease of use was the first criteria the software had to pass before the management team looked at "functionality" issues. Remember, the software portion of the system had to be simple enough for lab operators to use easily. They found RheoLogic much easier to use than the other systems they tested. With as little as three keystrokes operators can be running a test. The analysis portion of the system had to contain maximum flexibility.

With the ability to see more than ten batches at once, trends are much more obvious. Lab techs can now spot trends which may take 25 to 30 batches to become obvious. This critical information may have slipped by when looking at three or four ten-batch flat-bed charts. With this information the lab can notify the mixing department to request corrective action. If this is done before the trend goes out of spec, Hoover-Hanes can avoid mixing out of spec batches. This type of information can save thousands of dollars in scrap and rework. Valuable trend information was not very available with the old chart recorder sheets.

Another important benefit of the new system is that Hoover-Hanes can dispose of their overlays and set gates to individual compounds automatically. The lab can use much more data to set gate limits. They often use 100 batches to establish gate spread and location limits for new compounds or customers. Once gates are set, any time in the future this compound is run its gates are automatically on the screen. Specialized gates were simply not practical with the old system. The lab techs simply watch to make sure every curve passes within gates as they are drawn on the screen. They can see what's happening to each curve in real time, rather than wait for a completed flat-bed sheet. The system even tells the operator if a curve misses a gate.

Instead of filling filing cabinets with paper sheets, RheoLogic stores all rheology data in one central database. The other software systems store each compound's data in separate files. In these slower file-based systems, every compound has its own file name. The difference is important. Database systems are very fast and flexible, two attributes critical in the rubber industry.

Multiple sorting creates difficult problems for file-based rheology software systems. The lab manager at Hoover-Hanes likes to check how a certain compound ran given a specific set of production conditions. For instance, he may know that production will run this compound on mixer number 40, during third shift, using mixer operator Joe Smith. He can quickly pull up past runs of this compound matching this exact production set-up. If this set-up did not yield acceptable compounds, he can alter the new production set-up to produce an acceptable product. This type of multiple condition query often is not possible, or is very slow, with any file-based software systems.

There were other items Hoover-Hanes found undesirable about a file-based rheology system. This type of system yields hundreds of files. They create new files for each new compound number. A data sort (as in the above example) is very difficult and time consuming. A new file must be created for the compound only when it ran on mixer 40. From this new compound/mixer file another new file must be created from the previous file containing only the data from third shift. From the compound/mixer/shift file a newer file yet must be created sorting for operator Joe Smith. If you are counting, that's three files for one historical data search.

As this small example shows, file-based software systems easily become unwieldy to use and maintain. In addition, as more files accumulate the system runs slower and slower, and becomes increasingly difficult to manage.

The lab has a year and a half worth of data on disk storage in the computer. Every compound and batch is available quickly for historical analysis, customer audit, or just to check on the last results of that compound. Periodically, they run a compound that hasn't run for months or years. All of the historical rheology information is available in seconds.

All of the calculated data points, gate limits, customer specs, and so on are all there, on the screen. Using a file-based system it's difficult to find the correct file or append the correct set of files for the proper data. Data files in these systems can only get so large before they fill up. As a result, information on one frequently run compound may be in several separate files. To analyze data from these separate files you must create another file (by an appending process) from the appropriate bits and pieces of the others.

Finally, file based systems require much more time and computer hard disk storage space than a databased system. For example, the year and a half worth of data is only using less than one half of the 145 megabyte hard disk.

Statistical process control

SPC divides into two categories, process control and process improvement. Process control is determining if current product is in a state of statistical control. Basically, this means the batches currently in production are running as usual, with no assignable causes of variation affecting the quality. Process improvement is actively studying the process for assignable causes of poor or high quality in the production process. This means looking at historical data to find why certain production runs were bad or good. You learn from past production successes and failures with this information.

Many companies use control charts to help determine process control. A control chart is a graph of batch data plots for a given compound characteristic. A control chart plots these points over time in relation to upper and lower control limits. A batch plot above the upper control limit, or below the lower control limit, indicates some kind of manufacturing process change has occurred due to more than just chance. Most managers will investigate the manufacturing process for possible causes of this out of control situation.

Hoover-Hanes has control charts for tensile test results, elongation, modulus at 100%, specific gravity, Mooney viscosity and Mooney scorch. They may look at all six of these charts on one screen. There are many advantages to control charting. A control chart will tell you if a change in the process has occurred or is occurring. In addition, it will tell you how the process change is affecting the product (in a positive or negative manner). Finally, by showing batch plotting outside control limits, it will inform you when the process change is significant enough to justify action on your part.

For example, the lab manager noticed the control chart for ML had peaked above the upper control limit. This told him something had changed, and that this change was adversely affecting the product. After a manufacturing process investigation he found that a filter screen had clogged. Control charts are very sensitive to manufacturing process shifts. A lab will not catch these types of mistakes as soon, or at all, if they only check the torque curves of compounds. Control charts also apply to cure values. These charts make gradual trends (over more than 150 batches or more) very apparent. With this type of chart one can see if each successive batch is moving in a particular direction for say, MH.

Process improvement

Many rubber companies are trying to implement process improvement. The reason - process improvement saves money and improves the quality of their products. The problem is process improvement takes a lot of time and effort. Consider the following example:

TS2 (time to rise two inch-pounds above minimum torque) for a certain compound is unstable from batch to batch. Consistency is everything in producing high quality product. A process improvement study is in order. To begin the study, choose the compound's batch results at TS2 over many different runs or loads. Next, develop a histogram and bell curve from this data. Now, draw in the gates (or specs) at the correct locations.

Knowledge of the manufacturing process is critical in the next step. Why are the batches sometimes crowding one gate? Or, do most of the batches fall within gates but some arc above and below gate/spec limits. An investigation is in order. Many mangers will enlist help from plant supervisors and operators at this step. Usually one or two causes are most suspect. Now back to the data. Again, this is where a databased rheology software system is very effective. Imagine the mill number is the greatest suspected culprit. Now pull data on the compound one mill at a time. In this example there are three mills in which this compound ran over the last year. Develop three new histogram charts, one for each mill only when it ran this compound. By looking at these three separate histograms you may find that one mill produces batches that are on the low side of the gate limits. Another mill produces batches toward the high gate. The third mill produces batches in the center of gate limits. Congratulations! We have found the assignable cause of this problem. The mills are not producing consistent products.

By taking a direct measure of the mill temperature we discover all three of the mills are running at three different temperatures. This may be true despite the fact that all three read the same setting at the panels. The solution is to set each mill to the temperature which produced the best product; the third mill's actual temperature setting. Finally, we recalibrate all three temperature gages to read correctly.

As you can see, process improvement involves knowledge of the manufacturing process, and the data gathering required is extensive. A software system which is flexible to look at data in many different ways (in a reasonable amount of time) is essential. With the correct system, you will save considerable amounts of time, money and stand to dramatically improve quality in the long run.

Reporting

According to Hoover-Hanes, in the beginning, many of their customers were unfamiliar with the new RheoLogic computer reports. This made them nervous at first. They were used to getting seven flat-bed reports, and that was basically all the rheology information they had. Now, many customers insist on the RheoLogic charts and reports. The company sends an extensive 28 to 30-page report with each outgoing load. This detailed report consists of durometer, tensometer and Mooney viscometer results. In addition, they report specific gravity, and time to scorch (TS), time to cure (TC) calculated values. The packet begins with a numerical listing of all the data. Printouts of all the rheometer curves for each batch, and control charts with histograms for each physical property follow the raw data.

Immediately Hoover-Hanes experienced benefits of unchaining their company from the limitation of old technology. Although Hoover-Hanes sends a common report to all of its customers, the computerized system makes it easy to prepare custom reports for individual customers. One customer may not understand SPC and only wishes to see `standard rheology reports.' Other customers may request reports showing specialized data points.

Macros make the software system even easier to use. Macros automate tasks which require repeating keystrokes. A macro memorized the keystrokes, and will repeat them back in order, on command. For instance, Hoover-Hanes could set up a macro to print the 28 to 30-page report for them while they are at lunch. Some companies wish to make data entry extra simple for operators who do nothing in the system but run batch tests. A macro can "carry" the operator directly to the data entry portion of the software. All the operator needs to do is begin the test.

Conclusion

Hoover-Hanes recognized it could corporately benefit from plant automation, modernization and SPC before it was thrust upon them.

By purchasing cutting edge rubber testing equipment and software, they were able to effectively control and improve their production process. A major key to this improvement was tying lab data back into production. This could not be done with outdated technology. Information in the lab about the production process used for each batch (which mill, ingredient lot, operator, etc.) is critical for high levels of quality. The lab must have a method to see batch trends and batch to batch consistency. When the consistency fails they have to know why and communicate this back to production.

To stay competitive worldwide, rubber companies cannot keep repeating the same mistakes. We have to learn what the production process is telling us. Process improvement and process control are the one-two punch needed to increase quality and productivity. In the rubber industry, cutting edge rheology instruments and software are the necessary tools to begin down this path.

Mitch Mann, Paul Hertzler & Co.