Computational Modelling Group

Seminar  9th March 2011 1 p.m.  University of Southampton, Building 54, Room 10B

Conic programming formulations of some engineering problems

Dr Athanasios Makrodimopoulos
University of Southampton, School of Engineering Sciences

Web page
http://www.symbiosis.soton.ac.uk/index.php
Categories
Optimisation
Submitter
Petrina Butler

Numerical optimization is often coupled with numerical methods like FEM for the solution of various engineering problems. In order to obtain an accurate solution, a fine discretization may be needed. However this leads to optimization problems with a high number of variables and constraints. In most cases the solution is to avoid the use of general nonlinear algorithms, which may not be able to converge, and formulate the problem as a more special case where we can take advantage of features like the sparsity of the matrix data, the special form of constraints and the objective functions. Here we show that some engineering problems can be formulated as conic optimization problems (second – order cone programming, semi-definite programming). This is advantageous because such problems can be solved quickly and accurately even if they contain hundreds of thousands of constraints and variables, by using available interior point algorithms. Here two subjects are presented. The first is the limit and shakedown analysis of structures which is related to the calculation of the carrying capacity of elastoplastic structures. Structural optimization is the second topic and emphasis is given to truss optimization.

Contact

Dr Chris Marley

Research Fellow and Symbiosis Facilitator

School of Mathematics and Southampton Statistical Sciences Research Institute

cjm4v07@soton.ac.uk