Full text
| |
| Author(s): |
Daiane Gonçalves Ferreira
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Matemática, Estatística e Computação Científica |
| Defense date: | 2017-12-14 |
| Examining board members: |
Márcia Aparecida Gomes Ruggiero;
Kelly Cristina Poldi;
Aurelio Ribeiro Leite de Oliveira;
Douglas Soares Gonçalves;
Silvio Alexandre de Araújo
|
| Advisor: | Antonio Carlos Moretti; Márcia Aparecida Gomes Ruggiero |
| Abstract | |
The aim of this work is to address problems formulated as MINLP (Mixed Integer Nonlinear Programming). We propose a heuristic resolution method based on Inexact Restoration methods combined with the Feasibility Pump heuristic. The Inexact Restoration methods were proposed for solving nonlinear problems with continuous variables. These methods involve two phases, Restoration (viability phase) and Optimality. The Feasibility Pump heuristic was proposed to obtain feasible solutions for optimization problems with integer variables, MILPs (Mixed Integer Linear Programming) and MINLPs. In this work we adapt the two phases of the Inexact Restoration method in the context of problems with integer variables, MINLP, seeking advances in feasibility (Restoration phase) through the Feasibility Pump heuristic. In the optimality phase, two subproblems are solved, in the first the integrality constraints are relaxed and we construct a NLP (Nonlinear Programming), in the second the nonlinear constraints are relaxed and we construct a MILP. A master process coordinates the subproblems to be solved at each stage. The performance of the final algorithm was analised in a set of classical problems (AU) | |
| FAPESP's process: | 13/21515-9 - Models and algorithms for nonlinear mixed integer problems (MINLP) |
| Grantee: | Daiane Gonçalves Ferreira |
| Support Opportunities: | Scholarships in Brazil - Doctorate |