QUEST 2003-2005: Quality Engineering & Taguchi Methods
first part of the project required developing pseudo-codes for generating
resolution three and four asymmetrical orthogonal arrays with factors having
up to eight levels. Pseudo-codes were also developed for generating Morris
designs for computer experiments. The company essentially integrates design and
analysis of experiments tools with CAD/CAM for designing automobile components.
Toward this end, optimization of the engineering designs was enabled and
facilitated by providing and automating the tools of Taguchi methods, which
involved segregation of noise factors from control factors, designing inner
and outer array (computer) experiments using the codes developed in the first
part, and analysis and modeling of signal-to-noise (S/N) ratios and responses.
Standard regression model building techniques and other sophisticated methods
like radial basis functions were implemented in the analysis and modeling
module. These fitted models were further optimized using standard gradient
based methods and as well as genetic algorithms to finally yield optimal
engineering design parameters.