A fundamental solution method for inverse heat conduction problem (Q2486507)

The authors consider a 1D heat equation \(u_{t}-a^{2}u_{xx}=0\) posed in \( ( 0,1) \times ( 0,t_{\max }) \), with Robin type boundary conditions at 1: \(\beta \partial u/\partial x( 1,t) +\gamma u( 1,t) =g( t) \) and a Dirichlet boundary condition at some \(x^{\ast }\): \(u( x^{\ast },t) =h( t) \).\ The solution starts from some initial data \(\varphi \) at \(t=0\). The problem is to determine the temperature \(u( 0,t) \) and the flux \( \partial u/\partial x( 0,t) \). The authors use the fundamental solution as a radial basis function. They indeed look for a solution of the kind \(u( x,t) =\sum_{j=1}^{n+m+l}\lambda _{j}\phi ( x-x_{j},t-t_{j}) \) where the basis function is deduced from the fundamental solution \(F\) through \(\phi ( x,t) =F( x,t+T) \). The authors are thus faced with a linear system: \(\mathbf{A}\lambda =\mathbf{b}\), where \(\mathbf{A}\) is an \((n+m+l) \times (n+m+l) \) matrix. Because the original continuous problem is ill-posed, the matrix \(A\) is ill-conditioned. The authors introduce the Tikhonov regularisation of this problem, considering \(\min_{\lambda }( \| \mathbf{A}\lambda - \mathbf{b}\| ^{2}+\alpha ^{2}\| \lambda \| ^{2}) \), where \(\alpha \) is a parameter which is determined using the \( L\)-curve method. The main part of the paper presents some numerical results on a classical problem denoted as Beck's problem. They present curves describing some error between the exact and the approximate solutions.