scientific article; zbMATH DE number 3960939

From MaRDI portal
Publication:3729970





[For the entire collection see Zbl 0587.00014.] An inverse free boundary problem is defined as one in which the model that provides the theoretical values to be fitted to experimental data is given by a partial differential equation with unknown boundary conditions. Unlike the ordinary free boundary case the inverse free boundary problem requires the determination of the boundary condition while the boundary domain is known. Furthermore adequate representations of the unknown boundary conditions are to be obtained from the data in addition to the other parameters that enter into the model. The problem is shown to be ill posed in the sense of Hadamard, which means that representations of the unknown boundary conditions are to be searched in low dimensional spaces regardless of the number of data points available. The special case of linear partial differential equations is discussed and a general regression algorithm based on separable least squares and differentiation of pseudoinverses is proposed. Its application to the one dimensional time dependent diffusion equation with unknown interface reaction is examined in detail. The unknown boundary condition is described by piecewise time dependent polynomials and the continuity conditions across different time intervals provide additional constraints that, after substitution into the model, reduce the number of linear parameters considerably. A general strategy for a reasonable compromise between well-posedness and accuracy is also illustrated. A method for avoiding tedious algebraic manipulations connected with the elimination of both continuity constraints and linear parameters is presently being investigated using a newly developed constrained regression algorithm.











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 Q3729970)