New efficient techniques to solve sparse structured linear systems, with applications to truss topology optimization

Details

Supervisors

Glineur, François ; Nesterov, Yurii

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en mathématiques appliquées

Abstract

The subject of truss topology becoming a popular area of research, it considers the optimization of structures under different loading scenarios (external forces) and other conditions. Quality of such structures can be assessed by considering compliance as the main criterion, which we try to minimize while having an stable structure. Many formulations are derivable and lead to interesting resolution schemes. We focus our interest on the solving of huge-scale truss topology design problems whilst taking advantage of some of the interesting properties of such problems. The motivation behind this is that second order methods don't scale well when the problem becomes too big. We therefore investigate the use of first order methods, due to their cheap iteration cost, which could provide a more scalable approach, hopefully with realistic resolution times to solving large and sparse structures problems such as these