EXPERIMENTAL DESIGN AND ROBUST PARAMETER DESIGN IN MULTIPLE STAGE MANUFACTURING FOR NANO-ENABLED SURGICAL INSTRUMENTS
Open Access
- Author:
- Yuangyai, Chumpol
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 10, 2009
- Committee Members:
- Harriet Black Nembhard, Dissertation Advisor/Co-Advisor
Harriet Black Nembhard, Committee Chair/Co-Chair
Sanjay B Joshi, Committee Member
Dennis Kon Jin Lin, Committee Member
Mary I Frecker, Committee Member - Keywords:
- two-level fractional factorial design
split plot
split block
multistage experiments
minimum aberration - Abstract:
- The advent of rapid and exciting scientific advances in nanotechnology and nanomanufacturing allows scientists and engineers to create new and sophisticated products. However, the quality and yield of these products is still limited. Based on a review of the literature, we recognized several opportunities to use statistically-based design of experiments (DOE) and robust parameter design (RPD) in this field. More specifically, along with a team of Penn State researchers, we have been advancing a new multi-functional forceps-scissors (FS) instrument for minimally invasive surgery (MIS). The lost mold rapid infiltration forming (LMRIF) process is being developed to fabricate the tiny tool. There are many technical and quality issues that need to be overcome. Furthermore, when this novel process is established in the laboratory and ready to transition to full scale manufacturing, its continuing repeatability and reproducibility must be assured. In the experimentation to develop and refine the LMRIF process, there are restrictions on the randomization, by which we mean that the allocation of the experimental material and the order in which the individual trials of the experiment are to be performed are not randomly determined because certain process variables are “hard to change" or “expensive to change" due to the nature of the multiple stages that are involved in the process. Randomization, however, is one of the key principles in DOE. If the principle of randomization is violated, and the typical approach to data analysis is still employed, serious misinterpretation of the results may occur. This will lead to delays and/or diminished reliability in new product development. While there is a rich history and body of literature on DOE, there is a gap between the literature and the problem that we propose to address. In this research, we develop the multistage fractional factorial split-plot (MSFFSP) design with the combination of split-plot and split-block structure. Some properties are derived and its application is demonstrated in the green-bar yield improvement of the LMRIF process. Furthermore, we develop a framework of DOE and RPD to expedite the transition of micro- and nano-scale technologies into robust products that can be produced with minimum variability and defects. To maximize the information obtained from the MSFFSP design, we extend an algorithm from Bingham and Sitter (1999, 2001) to determine the optimal design under two general criteria: maximum resolution and minimum aberration. The algorithm is coded in MATLAB and is used to construct design catalogs for three and four stage experiments. An application to the LMRIF process is explored. In order to reduce the variability in the LMRIF process when it is transferred from the laboratory scale to full scale manufacturing, we integrate the MSFFSP design with the RPD concept. We focus on addressing multiple stages and multiple sets of noise factors in this integration, which is convenient for the LMRIF process. A foundation for using the concept is laid out and an optimal design catalog based on modified minimum aberration criteria for the variation reduction is provided for two-stage experiments with two sets of controllable factors and one set of noise factors. A computer code in MATLAB is also constructed for this purpose, and it can be used for larger experimentation. An application of the model is also explored for the improvement of the fired FS yield of the LMRIF process. The MSFFSP design and its integration with the RPD concept result in a more rapid understanding of the interaction among process conditions, product characteristics, and product reproducibility under constrained resources. This will not only advance the field of quality engineering in nanomanufacturing, but also has potential applications for other types of manufacturing.