We study the regularization problem for linear, constant coefficient descriptor systems $E x^. = AX + Bu, y_1 = Cx, y_2=\Gamma x^.$ by proportional and derivative mixed output feedback. Necessary and sufficient conditions are given, which guarantee that there exist output feedbacks such that the closed-loop system is regular, has index at most one and $E +BG\Gamma$ has a desired rank, i.e. there is a desired number of differential and algebraic equations. To resolve the freedom in the choice of the feedback matrices we then discuss how to obtain the desired regularizing feedback of minimum norm and show that this approach leads to useful results in the sense of robustness only if the rank of E is decreased. Numerical procedures are derived to construct the desired feedbacks gains. These numerical procedures are based on orthogonal matrix transformations which can be implemented in a numerically stable way.

We study the regularization problem for linear, constant coefficient descriptor systems $E x^. = AX + Bu, y_1 = Cx, y_2=\Gamma x^.$ by proportional and derivative mixed output feedback. Necessary and sufficient conditions are given, which guarantee that there exist output feedbacks such that the closed-loop system is regular, has index at most one and $E +BG\Gamma$ has a desired rank, i.e. there is a desired number of differential and algebraic equations. To resolve the freedom in the choice of the feedback matrices we then discuss how to obtain the desired regularizing feedback of minimum norm and show that this approach leads to useful results in the sense of robustness only if the rank of E is decreased. Numerical procedures are derived to construct the desired feedbacks gains. These numerical procedures are based on orthogonal matrix transformations which can be implemented in a numerically stable way.

[pt] As informações sobre a sensibilidade da solução de um dado escoamento mediante a perturbações infinitesimais é importante para o seu completo entendimento. A análise de estabilidade de escoamentos pode ser utilizada na otimização de processos industriais. Na indústria de revestimento o controle da estabilidade é fundamental, uma vez que o escoamento na região de aplicação da camada de líquido sobre o substrato, de um modo geral, tem que ser laminar, bidimensional e em regime permanente. O objetive é determinar, dentro do espaço de parâmetros de operação, a região onde o escoamento é estável e conseqüêntemente a camada a ser revestida uniforme. Porém, por ser uma análise complexa, só é usada na indústria em estudos mais apurados. O sistema linear que descreve a estabilidade vai ser discretizado com o método de Galerkin / elementos finitos, dando origem a um problema de autovalor generalizado.Tanto para escoamentos com líquidos newtonianos como para escoamentos com líquidos viscoelásticos, uma das matrizes do problema de autovalor generalizado é singular e alguns autovalores se encontram no infinito. No escoamento com líquidos viscoelásticos parte do espectro é contínuo, aumentando o grau de dificuldade da análise numérica para encontrá-lo. Nesse trabalho, vamos apresentar um método baseado em transformações lineares tirando vantagem das estruturas matriciais e transformando-as em um problema de autovalor clássico com dimens são, pelo menos, três vezes menor que o original. O método elimina os autovalores infinitos do problema com um baixo custo computacional. A estabilidade de um escoamento de Couette unidimensional de líquido newtoniano é analisada como um primeiro exemplo. Depois, o início do estudo da estabilidade em um escoamento de Couette bidimensional e também um escoamento pistonado com o mesmo líquido. Generaliza-se o método para o escoamento de Couette de um líquido viscoelástico, os resultados para o escoamento de um líquido cujo comportamento mecânico é descrito pelo modelo de Maxwell são apresentados e comparados com a solução analítica de Gorodtsov & Leonov, 1967. A relação entre os autovetores do problema transformado e do original é apresentada. / [en] Steady state,two-dimensional flows may become unstable under two and three-dimensional disturbances, if the flow parameters exceed some critical values. In many practical situations, determining the parameters at which the flow becomes unstable is essential. The complete understanding of viscous and viscoelastic flows requires not only the steady state solution of the governing equations, but also its sensitivity to small perturbations. Linear stability analysis leads to a generalized eigenvalue problem, GEVP. Solving the GEVP is challenging, even for Newtonian liquids, because the incompressibility constraint creates singularities that lead to nonphysical eigenvalues at infinity. For viscoelastic flows, the difficulties are even higher because of the continuous spectrum of eigenmodes associated with differential constitutive equations. The complexity and high computational cost of solving the GEVP have probably discouraged the use of linear stability analysis of incompressible flows as a general engineering tool for design and optimization. The Couette flow of UCM liquids has been used as a classical problem to address some of the important issues related to stability analysis of viscoelastic flows. The spectrum consists of two discrete eigenvalues and a continuous segment of eigenvalues with real part equal to -1/We (We is the Weissenberg number). Most of the numerical approximation of the spectrum of viscoelastic Couette flow presented in the literature were obtained using spectral expansions. The eigenvalues close to the continuous part of the spectrum show very slow convergence. In this work, the linear stability of Couette flow of a Newtonian and UCM liquids were studied using finite element method, which makes it easier to extend the analysis to complex flows. A new procedure to eliminate the eigenvalues at infinity from the GEVP that come from differential equations is also proposed. The procedure takes advantage of the structure of the matrices involved and avoids the computational effort of common mapping techniques. With the proposed procedure, the GEVP is transformed into a smaller simple EVP, making the computations more effcient. Reducing the computational memory and time. The relation between the eigenvector from the original problem and the reduced one is also presented.