Prediction of production yields in injection molding.

摘要

Plastics molders need to continuously improve production efficiency to remain competitive. With increasingly tight specifications driven by Six Sigma quality initiatives, such efficiency gains are more difficult to maintain. This thesis investigates the use of process capability indices for linear and non-linear regression models based on observed shrinkage data for polypropylene molded parts. The response surface models are developed by varying melt temperature, mold temperature, packing pressure, and injection rate using design of experiments. Yield estimates are provided by application of normal statistics to two individual length dimensions, their average, as well as, an equivalent part weight requirement. Statistical validation of the yield estimates is accomplished through Monte Carlo analysis. The discrepancy among the results indicates that current industry reliance on the univariate process capability index, Cp, is inadequate for parts with multiple critical dimensions, since it leads to errors in predicted yields and sub optimal process selection.