A monolithic approach to aero-structural analysis and optimization has been developed and implemented. In contrast to a partitioned approach which uses individual fluid and structural solvers to solve their respective systems separately, the monolithic approach solves a fully-coupled system simultaneously, enforcing solution compatibility across the sub-system interfaces at each iteration. In this work, a three-field formulation is used, consisting of fluid, structural, and fluid mesh-movement sub-systems. The performance of the monolithic approach is characterized using 1-D unsteady and 2-D steady analysis problems, and compared with a partitioned approach. Four steady model aero-structural optimization problems are also investigated. Gradients of the objective function are computed using the discrete-adjoint and flow-sensitivity (direct) methods. In each case, the monolithic approach is shown to be a promising option for efficient aero-structural analysis and optimization, though the implementation requires additional development of coupling sub-matrices when compared to a partitioned approach.

A monolithic approach to aero-structural analysis and optimization has been developed and implemented. In contrast to a partitioned approach which uses individual fluid and structural solvers to solve their respective systems separately, the monolithic approach solves a fully-coupled system simultaneously, enforcing solution compatibility across the sub-system interfaces at each iteration. In this work, a three-field formulation is used, consisting of fluid, structural, and fluid mesh-movement sub-systems. The performance of the monolithic approach is characterized using 1-D unsteady and 2-D steady analysis problems, and compared with a partitioned approach. Four steady model aero-structural optimization problems are also investigated. Gradients of the objective function are computed using the discrete-adjoint and flow-sensitivity (direct) methods. In each case, the monolithic approach is shown to be a promising option for efficient aero-structural analysis and optimization, though the implementation requires additional development of coupling sub-matrices when compared to a partitioned approach.

Uruguay is a pioneer in the use of renewable sources of energy and can usually satisfy its total demand from renewable sources. Control and optimization of the system is complicated by half of the installed power - wind and solar sources - be- ing non-controllable with high uncertainty and variability. In this work we present a novel optimization technique for efficient use of the production facilities. The dy- namical system is stochastic, and we deal with its non-Markovian dynamics through a Lagrangian relaxation. Continuous-time optimal control and value function are found from the solution to a sequence of Hamilton-Jacobi-Bellman partial differential equations associated with the system. We introduce a monotone scheme to avoid spurious oscillations in the numerical solution and apply the technique to a number of examples taken from the Uruguayan grid. We use parallelization and change of variables to reduce the computational times. Finally, we study the usefulness of extra system storage capacity offered by batteries.

Submitted by Maicon Juliano Schmidt (maicons) on 2015-07-15T14:25:14Z No. of bitstreams: 1 João Olegário de Oliveira de Souza.pdf: 4255697 bytes, checksum: 53d282b67d9c0e2f0b8d76e886a38a6f (MD5) / Made available in DSpace on 2015-07-15T14:25:14Z (GMT). No. of bitstreams: 1 João Olegário de Oliveira de Souza.pdf: 4255697 bytes, checksum: 53d282b67d9c0e2f0b8d76e886a38a6f (MD5) Previous issue date: 2013-01-31 / Nenhuma / Os controladores do tipo Proporcional, Integral e Derivativo, comumente denominados de PID, são largamente utilizados no controle de processos industriais, tanto em sistemas monovariáveis quanto em sistemas multivariáveis. Hoje, cerca de 95% dos controladores utilizam este tipo de estrutura na indústria. O grande problema é que grande parte deles estão mal sintonizados, comprometendo em muitos casos o desempenho de malhas industriais. Neste trabalho é apresentada uma revisão geral sobre os algoritmos inspirados na natureza, Simulated Annealing e Algoritmos Genéticos (fundamentos, características, parâmetros, operadores) e sua aplicação ao problema da sintonia de controladores PID monovariáveis e multivariáveis. É estabelecida, através de estudo de casos, uma análise comparativa entre estas sintonizações com metaheurísticas e os métodos consagrados na literatura em aplicações industriais convencionais, utilizando como função de avaliação o índice Integral do Erro Absoluto ponderado pelo Tempo (ITAE). O trabalho também propõe o estudo de controladores PID através de Algoritmos Genéticos Multiobjetivos, que satisfaçam dois critérios de desempenho: overshoot e o índice de desempenho Integral do Erro Quadrático ponderado pelo Tempo (ITSE). Conforme demonstrado pelos resultados obtidos, pode-se afirmar que a metaheurística Algoritmos Genéticos é um método eficiente e confiável para a otimização de problemas de sintonia de controladores PID. / The Proportional, Integral and Derivative controllers, commonly called PID controllers, are widely used in industrial process control, in both SISO and multivariable systems. Today about 95% of controllers use this type of structure in the industry. The big problem is that most of them are poorly tuned, in many cases compromising the performance of industrial loops. This work presents a general review on nature-inspired algorithms, Simulated Annealing and Genetic Algorithms (basement, characteristics, parameters, operators) and its application in the problem of tuning PID controllers in both single variable and multivariable systems. There will be through case studies, a comparative analysis of these metaheuristics with established methods in the literature in conventional industrial applications using as evaluation function the Integral of time multiplied by the Absolute Error (ITAE) index. The work also proposes the study of PID controllers using multiobjective genetic algorithms which meet two performance criteria: overshoot and the Integral Time Square Error (ITSE) index. The results obtained confirm that Genetic Algorithms are an effective and reliable method to optimize complex problems.