Abstract This thesis investigates a class of direct search methods for system optimization where it is assumed that the dependent variable is a unimodal function of N independent variables. The best optimization procedure is that which requires the least number of evaluations of the dependent variable. It is assumed that either the function or its derivative is unknown. The basic form of the search techniques studied requires the establishment of (N + 1) test points in N dimensions. Three...

Abstract The purpose of this paper is to investigate the suitability of dynamic simplex search methods for finding the global extremum of multimodal functions. The search will be in N variable space with the form of the function unknown or in such a form that conventional techniques such as the calculus cannot be used. The initial search is conducted using the dynamic simplex method to find the first extremum. Three methods were evaluated for their usefulness in determining the global...