scientific article; zbMATH DE number 1750408

From MaRDI portal
Publication:4533967





Starting from papers of \textit{H. W. Engl}, \textit{K. Kunisch} and \textit{A. Neubauer} [Inverse Probl. 5, 523--540 (1989; Zbl 0695.65037), \textit{A. Neubauer} [ibid. 5, 541--557 (1989; Zbl 0695.65038)] and \textit{T. I. Seidman} and \textit{C. R. Vogel} [ibid. 5, 227--238 (1989; Zbl 0691.35090)] the convergence theory for Tikhonov regularization of nonlinear ill-posed problem was developed by many authors and applied to a wide range of integral and differential equation problems. Here the authors present a new approach for obtaining convergence rates for Tikhonov regularization in the case of finding the temperature-dependent heat conductivity from temperature measurements by the nonlinear elliptic equation \[ -\nabla\cdot \bigl(q(u)\nabla u\bigr)= f(x) \text{ in }\Omega \subset\mathbb{R}^n \] with \(u(x)=0\) on \(\partial\Omega\). This approach avoids the usual strong source conditions and requirements of Fréchet-derivative smoothness of the parameter-to-solution mapping. Nevertheless, for noise level \(\delta\) a convergence rate \(O(\sqrt\delta)\) is obtained as obvious, where the additional assumption attains here the form of a regularity condition on the difference between the true parameter \(q^\dagger\) and the a-priori guess \(q^*\) used in the stabilizer of Tikhonov regularization. Such a condition seems to be more realistic than traditional source conditions for this identification problem. The method is not restricted to the one-dimensional case, but as case studies indicate a couple of numerical difficulties accompany the identification process also in one dimension.




Cited in
(16)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4533967)