Efficient algorithms for fractional factorial design generation
Abstract
In many industrial, engineering, and scientific fields, it is necessary to perform a sequence of tests (i.e., an experiment) to discover how various factors are related within some process. Since there are often a very large number of factors and exhaustive experiments would hence be infeasible, there is a critical need to uncover the most important effects using the fewest possible tests. Such a sequence of tests is called a "fractional factorial experimental design", since the test to be performed only encompass a fraction of the total tests associated with a full factorial design (abbreviated as FF with design omitted). Although fractional factorial design (FFD) is considered fundamental, it had not been efficiently mechanized. Further, various simplifications have been imposed (e.g., only considering sets of tests from a regular FFD), often resulting in the need to perform more tests than were necessary to obtain the desired information. The research attempts to automate generation of optimal regular FFDs (RFFDs) and irregular FFDs (IFFDs) for processes involving 2-level and 3-level factors. The FFDs are optimized to quantify specific information using the fewest tests.
Degree
Ph.D.
Advisors
Dietz, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.