DOE software eases experiment set-up.

Setting up a DOE, or "design of experiments," doesn't take as much know-how as it used to, thanks to software that can ease some of its more complex aspects. One such software package, Design-Ease from Minneapolis-based Stat-Ease Corp., has recently found use in several plastics applications, including injection molding, composites fabrication, and specialized manufacturing of ophthalmic lenses.

Design-Ease users say DOE methods fit well with a variety of plastics processes, which tend to have many variables and many interactions between those variables. Design-Ease, which runs on a Windows-based personal computer, generates "worksheets" that outline a set of processing experiments according to one of two DOE approaches - two-level factorial and Plackett-Burmann. It can handle up to 15 variables in two-level factorials and up to 31 variables with the Plackett-Burmann design. Following test runs on the production floor, Design-Ease can display the results in plots and tables - including analysis-of-variance (ANOVA) tables. Square and cube plots, for instance, show how key variables interact to affect the process [ILLUSTRATION FOR GRAPH OMITTED].

MANY VARIABLES AT ONCE

One of the reasons DOE works for plastics applications relates to the time it can save compared to experiments that vary a single process variable at a time. "We simply cannot afford to perform experiments in this slow manner," says Douglas Hubbel, process-improvement director for Los Angeles-based Anthony Industries, the maker of K-2 and Olin skis. So when optimizing its molding and lamination operations, the company turned to DOE techniques, which vary several variables at once according to mathematical guidelines.

This simultaneous look at the process variables can be a big time-saver. For example, Anthony has used Design-Ease to test 25 variables in as little as a few weeks - just a fraction what a one-variable-at-a-time approach would have taken, notes Hubbel. All the experimentation has paid off: Anthony has cut the reject rate from delamination from 1.8% before employing DOE to just 0.12% afterwards. And last year, the company's newly optimized molding operation won the Rochester Institute of Technology-USA Today Quality Cup.

More than just providing a faster route to experimentation, DOE shows the interactions between process variables. "By using a systematic approach to experimentation, not only can several process variables be considered at one time, but the degree of interaction between each variable can also be estimated. These capabilities make DOE a far more attractive approach than the traditional 'one-factor-at-a-time' approach," explains Richard Felton, senior process engineer at Sola Optical. In its Petaluma, Calif., lens plant, Sola has used Design-Ease for eight years on a variety of processes. Most recently, the company boosted the throughput of its lens coating line after DOE revealed optimum set-up parameters. Sola gained enough extra capacity, says Felton, to avoid a $250,000 equipment purchase despite increases in lens demand.

SIMPLIFYING MOLDING

DOE software has also found a place at a major injection molder. At its technical center in St. Petersburg, Fla., Tredegar Molded Products (recently purchased by Precise Technology) set up Design-Ease DOEs that have unearthed the sources of dimensional variations on a variety of jobs. According to project engineer Jay Carendar, Tredegar engineers typically use a two-level-factorial, 8- or 16-run DOE to study the effects of packing pressure, fill time, and mold temperature. "Interactions between the variables play a strong role in certain molds," says Carendar.

One such mold produces elliptical barrels for deodorant sticks. DOE showed that packing pressure and fill rate are the two key variables governing the elliptical sizing. "Raising packing pressure increased the dimensions of both axes," explains Carendar. "The fill rate mainly influences the relative size of the two axes. A faster fill changes the orientation of the plastic molecules in a way that increases shrinkage in the direction of flow. Increasing the fill rate also raises the temperature at which packing is accomplished." Tredegar made other DOE-driven findings, too: "We discovered that the effect of packing pressure on the part dimension becomes less at small gate sizes," Carendar adds.

With all the information revealed by DOE - presented in the form of graphic plots and ANOVA tables - Tredegar optimizes its process settings. "DOE provides a corrective road map on the shop floor," says Carendar. Reports generated by the DOE software follow jobs into the plant. That way, if a problem occurs, the machine operator can use the report to help solve it based on understanding which variables influence a given problem.

Carendar reports that it takes about one hour to come up with the experimental design and another 4-6 hr to mold a part using each of the prescribed combinations of variables. Testing parts takes another 4-6 hr. Finally, creating the report takes about another hour. Design-Ease costs $395.