Selection of Variables for Structural Redesign by Large Admissible Perturbations

Abstract

Redesign or inverse design is the process of generating a new optimal design which satisfies performance specifications starting from a baseline design with undesirable performance. The LargE Admissible Perturbation (LEAP) methodology makes it possible to redesign a structure for large changes in performance objectives and redesign variables without trial and error or repetitive finite element analyses. The next level of challenge in redesign automation is to identify a priori the structural elements and their properties that have the biggest impact and use only those in redesign. Based on LEAP, guidelines are developed in this paper for identifying effective selection of redesign variables for improved accuracy and reduced CPU time. These guidelines enable the designer to define the elements to be redesigned, to partition those elements among redesign groups, and to specify redesign variables in each group. In the numerical applications, an offshore tower is used to verify the developed guidelines. Three models of this tower with 160, 320, and 480 elements are used.