Further case studies are available on my Substack site:
https://effectivestatistics.substack.com/
Case Study: Injection Molding Process Optimization
The task was to develop a validated process for an injection molded part. The injection molding process had four process inputs, while the product had five characteristics (process outputs). The goal was for all five process outputs to show robust behavior to variation in process inputs, and to have a Cpk at or above 1.33.
As part of the Operational Qualification (“OQ” from IQ-OQ-PQ) we partnered with the process engineer to develop a Resolution IV Design of Experiments (DOE) study that incorporated the four process inputs to study how they, along with their interactions, affected the five process outputs.
The study incorporated blocking variables for cavity and station.
To augment the DOE, sample sizes and sampling procedures were created that would provide an early Phase 1 SPC chart of the process for the five outputs from the same data as the DOE.
The study was designed, the protocol was written and approved.
The DOE results provided information on statistically significant main effects and interactions. Output included ANOVA tables, interaction plots, and contour plots. Model assumptions were checked, including the behavior of model residuals.
Using the model from the DOE, settings were found for the four process inputs that provided stable behavior for all five process inputs, including robustness to anticipated variations in the process inputs.
Initially, the Cpk targets for four outputs were satisfactory while one output needed improvement. To address this, an Optimizer module was added, incorporating the DOE model. Weights were selected to place emphasis on the lagging output. The final result was a process, with parameter settings and windows, that was robust to variation in the four process inputs while providing acceptable process capability (Cpk) for the five process outputs.
EFFECTIVE STATISTICS 