Scientific injection molding does not include establishing process limits large enough to drive a truck through. The practice of using DOEs upfront to establish limits does not fall within scientific principles and should not be used for initial process development. A DOE is a problem-solving tool, why try and use it before a problem is identified?
Too often limits are established upfront using a handful of set points or generic outputs such as cycle time. The runs to determine these limits are often very short and don't allow enough time to truly prove that the process is repeatable at high limits and low limits.
Furthermore interactions are not accounted for. Trial runs are completed at the high end of the limits and then at the low end of the limits. But what about the mix in between? You could complete thousands of runs trying to account for every variable that these limits allow for. (And don't tell me that DOE software predicts this because with injection molding they are not effective.)
In the end all that has truly been created is a means for processing around abnormal conditions, which likely cause an unforeseen issue in the field or, if you're lucky, a defect caught before the product is released.
Upper and lower limits should be set using normal process variation during controlled runs. These controlled runs must be monitored closely to ensure no process changes are made. This data will show the true capability of the process that has been established. The goal is to monitor the established process outputs with limits specified using normal process variation.
Process limits should not allow for adjustments that will affect the established outputs to be made. For example: A world class injection molding machine should be able to maintain a fill time within ±0.02 second, as long as the process has been established using scientific injection molding principles ensuring the machine is and will never be pressure limited.
If the data that are collected over a 24-hour run prove that the fill time is maintained within this differential, the fill time limits should be set to ±0.03 seconds. This tight tolerance will ensure that normal or acceptable process variation does not cause unnecessary rejects. It also forces the processor to adjust the fill speed and transfer position to match the nominal output established during process development.
If your machines are not capable of running a ±0.02 second window, it is acceptable to increase the expectations, but anything over ±0.04 and I am calling a service rep for that machine manufacturer.
The point is that the typical methods of establishing process limits often cause more harm than good. These limits allow for process modification above and beyond what was originally intended. With the pressures put on processors to "just make it run", the temptation to make those modifications is hard to resist. Establishing tight effective monitoring limits will prevent these process changes from being made and help drive to true root causes.
About the Author: Robert P. Gattshall has worked 17 years in the automotive and medical injection molding industries, including 12 years in process engineering and process development. Certified in John Bozzelli's Scientific Injection Molding for more than 10 years, Gattshall has developed more than 600 processes using scientific injection molding principles. Certified in Lean 5S and SMED, Gattshall has also trained more than 50 process technicians and engineers on the principles of decoupled molding.