Mathematical modeling saves part inspection.

A new enhancement to a proprietary CIM system has allowed one injection molder to automatically predict part dimensions based on processing data. Security Plastics Inc. of Miami Lakes, Fla., implemented its mathematical modeling technique last January after two years of development. Since then, the molder reports that it has drastically reduced its dependence on manual inspection of parts, while increasing both the frequency and degree of automated inspection--looking at three part dimensions for each shot.

AN SPC ALTERNATIVE

Security Plastics, which operates 78 presses with multiple cavities per mold, has saved significantly in inspection costs by implementing mathematical modeling, according to CIM project engineer Michael Bentz. Conventional statistical process control (SPC) required sampling five parts every two to four hours and measuring critical dimensions of each part. Since implementing mathematical modeling, Security Plastics manually inspects parts only once a day to verify the mathematical modeling technique's accuracy.

The new approach is said to have inherent advantages over traditional "design-of-experiment" techniques which, through a series of prepared manual tests, establish "process windows" to determine the acceptability of part dimensions. Mathematical modeling does not calculate process windows or require manual tests or engineers to collect and analyze data. Rather, it uses a computer to automatically calculate the precise dimensions of each part before it leaves the mold.

Several benefits of this approach are claimed for the molder. It enables a process to be kept within acceptable limits before trending out of tolerance. Mathematical modeling provides information on where each part dimension lies on a curve, rather than just being termed "good" or "bad." Because it predicts part dimensions for each shot, it eliminates scrap that may occur between scheduled quality checks performed with SPC. The ability to detect abnormal variations and intermittent conditions has helped Security Plastics to achieve its goal of 6 sigma. Also, mathematical modeling is performed on a real-time basis on the plant floor; it therefore provides molders with immediate feedback after changing setpoints on a machine to adjust a process.

HOW IT WORKS

Mathematical modeling is based on the relationship between the mold, material, and molding machine. It combines calculations of key molding parameters such as mold temperatures on the fixed and moving side of the mold, pressures, screw position, and certain event times to predict the dimensional stability of the part. A key piece of hardware in the process is a Barber-Colman MACO 8000 controller, which has undergone certain enhancements to make possible an accurate model of the process.

According to Bentz, "Each process in each machine has a normal variation, and the part itself has a normal variation. In order to model the normal variation within the part, we had to model the normal variation in the process." Among the enhancements to the controller was the ability to read temperature to 0.045|degrees~ F, and screw position to three decimal places. This allowed the molder to see shot-to-shot variations in process parameters, and correlate them with shot-to-shot variations in part dimensions.

CREATING THE MODEL

When a mold first goes into production, a mathematical model is created based on manual inspection of the first 100 parts. The part dimensions are correlated to relevant process data recorded by the MACO 8000, and a formula is created that best describes these relations. The mathematical model is downloaded to the injection machine controller, where it runs independently of the host computer. Creating a mathematical model takes about 15 min. Once in production, the mathematical model takes about 1 sec to predict part dimensions based on the actual process parameters. The prediction also takes into account warpage and shrinkage that would take place after the part is molded.

When a tool is taken out of production and placed back in service on another press, the mathematical model is reloaded into the machine controller and five parts are sampled. If measured dimensions do not correlate with predicted ones, the model is updated on a computer to accommodate the characteristics of the new press, and again downloaded to the machine.

CLOSING THE LOOP

Now that Security Plastics has completed the development and implementation of mathematical modeling, to plans to enter a second phase of what it calls "closing the loop." This means that the control not only will calculate the part dimensions before the mold opens, but automatically adjust the machine operation to stay within acceptable tolerances, according to Bentz.