A widely convergent method for determining the distributed parameters of the elliptical equation (Q1316121): Difference between revisions

The authors study identification problems in elliptic partial differential equations. More precisely, the following equation defined on the time-space domain is considered \(L(a(x),t)u(x,t) = 0\) where \(a(x_ 1,\dots,x_ p)\) is the unknown operator to be identified. Of course initial and boundary conditions are given: \(Eu(x,0) = g(x)\); \(Bu(x,t) = f(x,t)\) on the boundary; \(Au(x,t) = \varphi(x,t)\) on a part of the boundary. A nonlinear operator \(Q\) is introduced: \(Q(a,t) = Au(a;x,t) - \varphi(x,t)\) and the inverse problem is stated as the following nonlinear operator equation: (1) \(Q(a,t) = 0\). The principle of continuation homotopy is used for solving (1) associated to a Newton-like method. Numerical results are given. Reviewer's remark: For this kind of problem I suggest to use a decomposition method (Adomian's technique) allowing to obtain \(u(x,t)\) in function of \(a(x)\) explicitly.