Showing papers on "Local search (optimization) published in 1971"
TL;DR: In this paper, the constrained optimization problem is transformed into a single unconstrained problem through the use of penalty functions constructed so as to create a reasonably symmetrical ridge at the acceptable-unacceptable region boundary of the composite merit surface.
Abstract: The constrained optimization problem usually encountered in automated design is transformed into a single unconstrained problem through the use of penalty functions constructed so as to create a reasonably symmetrical ridge at the acceptable-unacceptable region boundary of the composite merit surface. A variable step direct search, coupled with a more extensive local search at points of direct search failure, is used to move to, and then along, this ridge. The application of this algorithm to a series of integrally stiffened cylindrical shell studies, including shells with spiral-type stiffeners, indicates that it has good convergence properties and computational efficiency.
25 citations
01 Dec 1971
TL;DR: A new automated procedure for the computer-aided design of linear, time-invariant, multivariable control systems is developed for the case where the design specifications are given in the frequency domain in the form of stability margins, loop gains, and closed-loop frequency responses.
Abstract: A new automated procedure for the computer-aided design of linear, time-invariant, multivariable control systems is developed for the case where the design specifications are given in the frequency domain in the form of stability margins, loop gains, and closed-loop frequency responses. The procedure minimizes a cost (penalty) function which measures the deviation of the system responses from those specified to obtain the required fixed-configuration compensation to meet the specifications. The procedure is applied to a representative aircraft control system design problem using three local search techniques (optimium step-size gradient, Fletcher-Powell, and Rosenbrock pattern search) and a new "global" random search technique for the required minimizations. The new "global" random search technique appears to be superior to the local search techniques for the class of problems studied.
3 citations