| This thesis investigated a process optimization technique (for the injection molding process) based on design of experiments and numerical optimization. Traditionally, molders have used the trial and error approach for process set up. However, with the advances in current computer software and hardware, "optimizing" the process to satisfy multiple quality and cost objectives has become feasible. This thesis validates a process optimization method including: (1) design and implementation of a design of experiments, (2) developing multiple regression models for the responses, and (3) numerical optimization of stated objectives.; The implemented optimization methods were able to produce parts having a range of lengths, without flash or short shots, of 159.142mm--161.652mm for a tensile bar, and 123.306mm--124.577mm for a flex bar. They also produced parts with an average thickness of 3.251mm and at a lower value of 3.175 while minimizing the cycle time. When comparing different regression models, the results indicated that the MLR (Multiple Linear Regression) model optimized the process better than the PLS (Partial Least Squares) model.; The major drawback to this optimization method is that a sizeable design of experiments is required to develop statistically valid regression models. For low production processes with many factors, the upfront implementation costs may not yield a significant return on investment compared to the traditional approaches. |
