Integration et optimisation multidisciplinaire d'une turbine a gaz au stade preliminaire de conception (French text)

This master's thesis present the integration and Multidisciplinary Design Optimization (MDO) of a gas turbine problem using three disciplines: performance, aerodynamics and rotating structures. The integration is done using wrappers written in Perl language.; The MDO is conducted using the iSIGHT software and a Sequential Quadratic Programming method (SQP) named NLPQL.; The resolution of Multi-Disciplinary Analysis (MDA) problem has to be done in this work because there is a coupling between the disciplines studied. Two single level strategies are studied to resolve the MDA problem, in order to compare their performances. The first one, named Fully Integrated Optimization (FIO), solve the MDA problem each time the optimizer need a turbine analysis. The second one, named Distributed Analysis Optimization (DAO), introduce additional constraints to relax the coupling and avoid the need to solve the MDA problem at each turbine analysis.; Results show the potential of the DAO strategy to give a converged optimal solution in less evaluation calls and execution time then the FIO strategy. Moreover, it can be seen that a converged DAO optimization solution could be obtained in less than a minute if gradients computation were parallelized on a distributed network.