A spline-based parameter estimation technique for static models of elastic structures (Q1121698)

The inverse problem is considered in the paper. Spatially varying coefficient of elasticity e(x,y) is identified using measured u(x,y) and f(x,y). The observed solution u(x,y), according to the equation \((\partial /\partial x)(e\partial u/\partial x)+(\partial /\partial y)(e\partial u/\partial y)=f\), is the vertical displacement of the plane structure while f(x,y) is the distributed load. It is shown that under certain circumstances the mapping from \(f\in L^ p\) to \(e\in L^ p\) is continuous for all \(p\in (1,\infty)\). Then this continuous dependence in \(L^ 2(\Omega)\) is used to propose a spline-based technique for approximating the unknown coefficient e in \(\nabla e\cdot \nabla u+e\Delta u=f\), \(x\in \Omega \subseteq R^ n\). It is proved that the proposed scheme converges to the actual solution e and the error estimate is obtained. The discussion is supported by an example solved for a unit square domain.