. The identification of discontinuous parameters in elliptic systems is formulated as a constrained minimization problem combining the output least squares and the equation error method. The minimization problem is then proved to be equivalent to the saddle-point problem of an augmented Lagrangian. The finite element method is used to discretize the saddle-point problem, and the convergence of the discretization is also proved. Finally, an Uzawa algorithm is suggested for solving the discrete saddle-point problem and is shown to be globally convergent. Key words. parameter identification, elliptic system, augmented Lagrangian, finite element method AMS subject classifications. 65N30, 35R30 PII. S0363012997318602 1. Introduction. The main purpose of this paper is to propose a numerical approach and conduct convergence analyses on each approximation process in the identification of the unknown coe#cient q in the elliptic problem -# (q#u) = f in ## u = 0 on #. The identifying process i...

In this paper a family of trust-region interior-point SQP algorithms for the solution of a class of minimization problems with nonlinear equality constraints and simple bounds on some of the variables is described and analyzed. Such nonlinear programs arise e.g. from the discretization of optimal control problems. The algorithms treat states and controls as independent variables. They are designed to take advantage of the structure of the problem. In particular they do not rely on matrix factorizations of the linearized constraints, but use solutions of the linearized state equation and the adjoint equation. They are well suited for large scale problems arising from optimal control problems governed by partial differential equations. The algorithms keep strict feasibility with respect to the bound constraints by using an affine scaling method proposed for a different class of problems by Coleman and Li and they exploit trust--region techniques for equality-constrained optimizatio...

Abstract not found

We introduce an output least-squares method for impedance tomography problems that have regions of high conductivity surrounded by regions of lower conductivity. The high conductivity is modeled on network approximation results from an asymptotic analysis and its recovery is based on this model. The smoothly varying part of the conductivity is recovered by a linearization process as is usual. We present the results of several numerical experiments that illustrate

The problem of recovering a parameter function based on measurements of solutions of a system of partial differential equations in several space variables leads to a number of computational challenges. Upon discretization of a regularized formulation a large, sparse constrained optimization problem is obtained. Typically in the literature, the constraints are eliminated and the resulting unconstrained formulation is solved by some variant of Newton's method, usually the Gauss-Newton method. A preconditioned conjugate gradient algorithm is applied at each iteration for the resulting reduced Hessian system. In this paper we apply instead a preconditioned Krylov method directly to the KKT system arising from a Newton-type method for the constrained formulation (an "all-at-once" approach). A variant of symmetric QMR is employed, and an effective preconditioner is obtained by solving the reduced Hessian system approximately. Since the reduced Hessian system presents significa...

. We propose several formulations for recovering discontinuous coefficients in elliptic problems by using total variation (TV) regularization. The motivation for using TV is its well-established ability to recover sharp discontinuities. We employ an augmented Lagrangian variational formulation for solving the output-least-squares inverse problem. In addition to the basic output-least-squares formulation, we introduce two new techniques to handle large observation errors. First, we use a filtering step to remove as much of the observation error as possible. Second, we introduce two extensions of the output-least-squares model; one model employs observations of the gradient of the state variable while the other uses the flux. Numerical experiments indicate that the combination of these two techniques enables us to successfully recover discontinuous coefficients even under large observation errors. 1. Introduction. Consider the partial differential equation ae \Gammar \Delta (q(x)ru) =...

We try to give a brief survey about using multiple level set methods for identifying piecewise constant or piecewise smooth functions. A general framework is presented. Application using this general framework for different practical problems are shown. We try to show some details in applying the general approach for applications to: image segmentation, optimal shape design, elliptic inverse coefficient identification, electricall impedance tomography and positron emission tomography. Numerical experiments are also presented for some of the problems.

: Estimation of coefficients of partial differential equations is ill-posed. Output-least-squares method is often used in practice. Convergence of the commonly used minimization algorithms for the inverse problem is often very slow. By using the augmented Lagrangian method, the inverse problem is reduced to a coupled linear algebraic system, which can be solved efficiently. Total variation techniques have been successfully used in image processing. Here, we use it with the augmented Lagrangian approach to recover discontinuous coefficients. The numerical results show that our approach can recover discontinuous coefficients with large jumps from noisy observations. 1. Introduction Consider the partial differential equation: ae \Gammar \Delta (qru) = f; on\Omega ; u = 0 ; on @\Omega : (1) Our concern is that we know an approximate observation u d for u, i.e. u d ß u, and need to recover the coefficient q. A common approach is the output-least-squares method, i.e. find the minimize...

. We propose several formulations for recovering discontinous coefficient of elliptic problems by using total variation (TV) regularization. The motivation for using TV is its well-established ability to recover sharp discontinuities. We employ an augmented Lagrangian variational formulation for solving the output-least-squares inverse problem. In addition to the basic output-least-squares formulation, we introduce two new techniques to handle large observation errors. First, we use a filtering step to remove as much of the observation error as possible. Second, we introduce two extensions of the output-least-squares model; one model employs observations of the gradient of the state variable while the other utilizes the flux. Numerical experiments indicate that the combination of these two techniques enables us to successfully recover highly discontinous coefficient even under observation errors as high as 100% in the L 2 norm. 1. Introduction. Consider the partial differential equa...

Introduction This work continues our earlier investigations [2], [3] and [8]. The intention is to develop efficient numerical solvers to recover the diffusion coefficient, using observations of the solution u, from the elliptic equation \Gammar \Delta (q(x)ru) = f(x) in\Omega ; u = 0 on @\Omega : (1) Our emphasis is on the numerical treatment of discontinuous coefficient and efficiency of the numerical methods. It is well known that such an inverse problem is illposed. Its numerical solution often suffers from undesirable numerical oscillation and very slow convergence. When the coefficient is smooth, successful numerical methods have been developed in [5] [7]. When the coefficient has large jumps, the numerical problem is much m