The decomposition approach to inverse heat conduction (Q1284511): Difference between revisions

The authors deal with the one-dimensional inverse problem concerning the heat equation \[ u_{xx}- u_t= 0,\quad 0< x< 1,\quad t>0, \] in which the temperature \(u\) and the heat flux histories \(f_0(t)\) and \(g_0(t)\) on the left boundary \(x=0\) are to be determined from measurements \(f_1(t)= u(1,t)\) and \(g_1(t)= u_x(1,t)\) at the right boundary \(x=1\). This inverse boundary value problem in heat conduction is solved by using Adomian's decomposition approach. The considerations are based on the general representation of the solution: \[ u(x,t)= \Biggl\{\sum^\infty_{n= 0}{(x- 1)^{2n}\over (2n!)} f^{(n)}_1(t)+ {(x- 1)^{2n+ 1}\over (2n+ 1)!} g^{(n)}_1(t)\Biggr\}. \] Since the solution is instable for noisy data \(f_1\) and \(g_1\), a mollifying procedure \(J_\sigma\) is applied to that approach. Then it can be shown that the mollified solutions \(J_\sigma f_0\) and \(J_\sigma g_0\) converge to the exact solutions with a rate \(O([\ln(\varepsilon^{-1/2})]^{- 1})+ O(\varepsilon^{1/2})\) for an appropriate choice of \(\delta\) if the noise level \(\varepsilon\) for the data \(f_1\) and \(g_1\) tends to zero. Some variants of the inverse problem, for example the mixed boundary value problem, are also considered based on Adomian's decomposition method.