METHODS OF PROCESS OPTIMIZATION
Date
1964-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
Type
Degree Level
Doctoral
Abstract
The theory of two level factorial designs is derived by replacing the continuous independent variables of an infinite power series with binary variables. The analysis of unrepli-cated designs is then described in terms of a Hadamard matrix with particular attention to a proposal for their use in automatic process optimization. Using the matrix, the theory of regular fractional factorials is developed and it is shown how these designs can be automatically generated by a digital computer. Algorithms based upon logical multiplication are derived which permit specification of some confoundings during the design procedures.
Digital simulations are used to demonstrate the theor-etical points and then to compare experimentally several different process optimizing methods which have been pro-posed in the literature by measuring the time integrated errors which they accumulate under similar operating condi-tions. Fractional factorial experiments, automatically designed and analyzed, are used to estimate process deriva-tives in those methods which require it. It is shown that the method which gathers the most information is not neces-sarily the best optimizer in the sense that it does not minimize the time integrated error. Systems are compared under noiseless as well as noisy conditions at the process output.
Description
Keywords
Citation
Degree
Doctor of Philosophy (Ph.D.)
Department
Electrical Engineering