Hierarchical adaptive multi-objective fuzzy optimization and applications to circuit design and IC fabrication process simulation

A new methodology based on the use of Fuzzy Sets Theory and intended to be applied to the design of products or processes having a number of requirements is outlined. It was initially motivated by the intention to generalize the Taguchi methodology to an arbitrary number of outputs and statistical measures as well as to allow the user to specify an ordering of the requirements to be followed during the optimization process. It relies heavily on the notions of dynamically altered membership functions and aggregation procedures. It also makes strong usage of history information to guide both the search procedure in the optimization space as well as to control the dynamically altered Membership functions and the way they are aggregated. The methodology is applicable to generic situations where the circuit or process under study must obey a number of criteria of various natures, which can but do not need to be of a statistical type. Included applications of the methodology are: (1) a CMOS OPAMP; (2) the segment of a NMOS transistor fabrication process simulation and (3) a couple of highly non-linear test vector functions.