Band 16, Heft 2

The estimation of distributed parameters in a partial differential equation (PDE) from measures of the solution of the PDE may lead to underdetermination problems. The choice of a parameterization is a frequently used way of adding a priori information by reducing the number of unknowns according to the physics of the problem. The refinement indicators algorithm provides a fruitful adaptive parameterization technique that parsimoniously opens the degrees of freedom in an iterative way. We present a new general form of the refinement indicators algorithm that is applicable to the estimation of distributed multidimensional parameters in any PDE. In the linear case, we state the relationship between the refinement indicator and the decrease of the usual least-squares data misfit objective function. We give numerical results in the simple case of the identity model, and this application reveals the refinement indicators algorithm as an image segmentation technique .

We consider the operator H := i∂ t + ∇ . ( c ∇) in an unbounded strip Ω in ℝ 2 , where . We prove an adapted global Carleman estimate and an energy estimate for this operator. Using these estimates, we give a stability result for the diffusion coefficient c ( x , y ).

We recover locally in time two (smooth) unknown convolution kernels, depending on time only, in a phase-field system coupling two hyperbolic integro-differential equations, where the differential operators (of order two and four in space, respectively) appear only under the integral sign. Moreover, we can prove a global uniqueness result concerning the unknown kernels.

We propose a new approach to weighting initial parameter misfits in a least squares optimization problem for linear parameter estimation. Parameter misfit weights are found by solving an optimization problem which ensures the penalty function has the properties of a χ 2 random variable with n degrees of freedom, where n is the number of data. This approach differs from others in that weights found by the proposed algorithm vary along a diagonal matrix rather than remain constant. In addition, it is assumed that data and parameters are random, but not necessarily normally distributed. The proposed algorithm successfully solved three benchmark problems, one with discontinuous solutions. Solutions from a more idealized discontinuous problem show that the algorithm can successfully weight initial parameter misfits even though the two-norm typically smoothes solutions. For all test problems sample solutions show that results from the proposed algorithm can be better than those found using the L-curve and generalized cross-validation. In the cases where the parameter estimates are not as accurate, their corresponding standard deviations or error bounds correctly identify their uncertainty.

A class of approximate methods to solve operator equations of first kind with not exactly given input data is constructed. For involved methods their optimality by the order on sets of sourcewise represented solutions is proved and the bound of informational expenses is obtained. These algorithms are numerically implemented in an efficient way. An example of application of two such algorithms is given.