This dissertation deals with frequency domain identification of linear dynamic systems in a deterministic set-up. Various untuned algorithms are suggested, including one which is robustly convergent and asymptotically optimal (in n-width sense) for a finite model order. The suggested algorithms can easily be implemented in commercially available software for convex optimization.

Model reduction is a powerful and ubiquitous tool used to reduce the complexity of a dynamical system while preserving the input-output behavior. It has been applied throughout many different disciplines, including controls, fluid and structural dynamics. Model reduction via proper orthogonal decomposition (POD) is utilized for of control of partial differential equations. In this thesis, the analytical expressions of POD modes are derived for the heat equation. The autocorrelation function of the latter is viewed as the kernel of a self adjoint compact operator, and the POD modes and corresponding eigenvalues are computed by solving homogeneous integral equations of the second kind. The computed POD modes are compared to the modes obtained from snapshots for both the one-dimensional and two-dimensional heat equation. Boundary feedback control is obtained through reduced-order POD models of the heat equation and the effectiveness of reduced-order control is compared to the full-order control. Moreover, the explicit computation of the POD modes and eigenvalues are shown to allow the computation of different n-widths approximations for the heat equation, including the linear, Kolmogorov, Gelfand, and Bernstein n-widths.

Les méthodes de réduction de modèles offrent un cadre général permettant une réduction de coûts de calculs substantielle pour les simulations numériques. Dans cette thèse, nous proposons d’étendre le domaine d’application de ces méthodes. Le point commun des sujets discutés est la tentative de dépasser le cadre standard «bases réduites» linéaires, qui ne traite que les cas où les variétés solutions ont une petite épaisseur de Kolmogorov. Nous verrons comment tronquer, translater, tourner, étirer, comprimer etc. puis recombiner les solutions, peut parfois permettre de contourner le problème qui se pose lorsque cette épaisseur de Kolmogorov n’est pas petite. Nous évoquerons aussi le besoin de méthodes de stabilisation sur-mesure pour le cadre réduit. / Model reduction methods provide a general framework for substantially reducing computational costs of numerical simulations. In this thesis, we propose to extend the scope of these methods. The common point of the topics discussed here is the attempt to go beyond the standard linear "reduced basis" framework, which only deals with cases where the solution manifold have a small Kolmogorov width. We shall see how truncate, translate, rotate, stretch, compress etc. and then recombine the solutions, can sometimes help to overcome the problem when this Kolmogorov width is not small. We will also discuss the need for tailor-made stabilisation methods for the reduced frame.