The influence of the equation type on iterative parameter identification problems which are elliptic or hyperbolic in the parameter (Q4433291)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 1998662
Language Label Description Also known as
default for all languages
No label defined
    English
    The influence of the equation type on iterative parameter identification problems which are elliptic or hyperbolic in the parameter
    scientific article; zbMATH DE number 1998662

      Statements

      The influence of the equation type on iterative parameter identification problems which are elliptic or hyperbolic in the parameter (English)
      0 references
      0 references
      0 references
      2003
      0 references
      Least squares method
      0 references
      error analysis
      0 references
      Landweber iteration
      0 references
      design of car windscreen
      0 references
      elastic plate
      0 references
      existence
      0 references
      weak solution
      0 references
      The equation modelling the design of car windscreens is related to the determination of Young's modulus \(E\) in the following linear elastic plate problem, where \(w\) denotes the displacement: \[ \begin{multlined} D_x^2[E(D_x^2w+\nu D_y^2w)]+ 2(1-\nu)D_xD_y (ED_xD_yw)+D_y^2 [E(D_y^2w+\nu D_x^2w)]\\ =-12h^{-3}(1-\nu^2)f,\quad \text{in\;} \Omega, \end{multlined}\tag{1} \] \[ w=g,\quad \text{ on\;} \partial \Omega, \tag{2} \] plus an appropriate boundary condition. Instead of (1) the authors deal with the simplified second-order problem \[ - D_x(ED_xw) - D_y(ED_yw) - D_x(wD_xE) - D_y^2E = f,\quad \text{ in\;} \Omega, \tag{3} \] \[ w=g,\quad \text{ on\;} \partial \Omega, \tag{4} \] preserving the second-order dependence on \(E\) as the original equation (1). Denoting by \(w(E)\) the solution of (3), (4), the inverse problem related to (3), (4) consists of determining \(E\) as the minimizer of the functional \(J(E)=\| w(E)-{\widetilde w}\| _{L^2(\Omega)}\), where \({\widetilde w}=g\) on \(\partial \Omega\) and \({\widetilde w}\) is affected by a known error \(\delta\): \(\| {\widetilde w}-{\widetilde w}_\delta\| _{L^2(\Omega)}\leq \delta\). Assuming, for the time being, that \(w={\widetilde w}\) is \textit{known}, from (3) we deduce the second-order equation for \(E\) \[ - D_x({\widetilde w}D_xE) - D_y^2E - D_x(ED_x{\widetilde w}) - D_y(ED_y{\widetilde w}) = f,\quad \text{ in\;} \Omega, \tag{5} \] Such an equation -- in general of mixed type -- is elliptic in \(\Omega_+=\{(x,y)\in \Omega: {\widetilde w}(x,y)>0 \}\) and hyperbolic in \(\Omega_-=\{(x,y)\in \Omega: {\widetilde w}(x,y)<0 \}\). First the authors state an existence results for the weak solution to the direct problem (3), (4), whenever \(E\in K=\{e\in H^1(\Omega): {\underline E}\leq e(x,y)\leq {\overline E},\;(x,y)\in \Omega,\;D_xe\in L^\infty(\Omega)\}\) and \(0\) does not belong to the spectrum \(\Sigma(E)\) of the linear operator \(L=- D_x(ED_x) - D_y(ED_y) - D_y^2 - (D_xE)D_x - (D^2_xE)\cdot\). To solve their inverse problem the authors choose the Landweber iteration method that in the present case consists in: {(i)} determining the solution \(w(E_{k-1})\) to problem (3), (4) with a given \(E_{k-1}\); {(ii)} determining the solution \(u_{k-1}\in H^1_0(\Omega)\) to the adjoint problem \[ - D_x(E_{k-1}D_xu_{k-1}) - D_y(E_{k-1}D_yu_{k-1}) - D_x(u_{k-1}D_xE_{k-1}) = w(E_{k-1})-{\widetilde w}; \tag{6} \] {(iii)} solving the following problem \[ {\overline E}_k - \Delta {\overline E}_k = - \nabla u_k \cdot \nabla D_xw_{k-1} + D_x(w_{k-1}D_xu_{k-1}) - D_y^2u_{k-1},\quad \text{ in\;} \Omega,\quad {\overline E}_k \in H^1_0(\Omega); \tag{7} \] {(iv)} setting \(E_k=E_{k-1}+{\overline E}_k\) and iterating the procedure. In the numerical experiments -- with both exact and noisy data -- the authors assume \(E=b\) on \(\partial \Omega\) and are concerned with how the change of type in equation (5) may affect the minimization result, although equation (5) is not directly involved in the iteration procedure. Further, three inverse problems related to equation (5), whose explicit solutions are known for specific \(f\)'s and \(g\)'s, are tested. The first one is elliptic in \(E\), the second one is hyperbolic, while the third one is of mixed type. In this case a comparative analysis of the relative and absolute errors is carried out. Finally, also an analysis of small perturbation on the boundary data is dealt with.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references